Merge branch 'dhk-1' of http://192.168.0.200/knowledge-base-bj/deerflow-code into dhk-1

Resolve conflicts by accepting master version for:
- workspace-nav-chat-list.tsx
- ResearchSidebar.tsx

Co-authored-by: Cursor <cursoragent@cursor.com>

.gitignore

@@ -0,0 +1,191 @@
+# Workspace/editor
+.vscode/*
+!.vscode/launch.json
+!.vscode/tasks.json
+!.vscode/extensions.json
+!.vscode/settings.json
+frontend-web/.vscode/
+node_modules/
+# Claude Code local-only settings (per-user permissions, not shared)
+.claude/settings.local.json
+
+# Local runtime logs and pid files
+.runtime/
+logs/
+offline-backend-20260512/backend/logs/
+
+# Frontend local/runtime artifacts
+frontend-web/.env
+frontend-web/.env.local
+frontend-web/.env.*.local
+frontend-web/node_modules/
+frontend-web/.pnpm-store/
+frontend-web/tsconfig.tsbuildinfo
+frontend-web/vite-dev.log
+frontend-web/vite-dev.err.log
+
+# Frontend local packages/archives/docs (non-source)
+frontend-web/pnpm-lock.yaml
+frontend-web/README.md
+frontend-web/doc.md
+frontend-web/src.zip
+frontend-web/src9.zip
+frontend-web/src-v1.zip
+frontend-web/src-*.rar
+frontend-web/dist
+# Frontend build artifacts
+frontend-web/dist/
+frontend-web/dist-scheduled-check/
+
+# Offline backend bundles and local deployment artifacts
+offline-backend-20260505.zip
+offline-backend-20260505/output/
+offline-backend-20260505/bundle/.env
+offline-backend-20260505/bundle/.env.example
+offline-backend-20260505/bundle/DEPLOY_BACKEND_OFFLINE_ZH.md
+offline-backend-20260505/bundle/SHA256SUMS
+offline-backend-20260505/bundle/config.yaml
+offline-backend-20260505/bundle/config.yaml.example
+offline-backend-20260505/bundle/data/
+offline-backend-20260505/bundle/deerflow-backend-offline-image-20260505.tar
+offline-backend-20260505/bundle/extensions_config.json
+offline-backend-20260505/bundle/frontend-vite.zip
+offline-backend-20260505/bundle/skills/
+offline-backend-20260505/bundle/start_backend.0509.sh
+offline-backend-20260505/bundle/start_backend.sh
+offline-backend-20260505/bundle/stop_backend.sh
+
+# offline-backend-20260512 runtime artifacts (large files, local config, data)
+offline-backend-20260512/.env.example
+offline-backend-20260512/.env
+offline-backend-20260512/DEPLOY_BACKEND_OFFLINE_ZH.md
+offline-backend-20260512/SHA256SUMS
+offline-backend-20260512/config.yaml.example
+offline-backend-20260512/data/
+offline-backend-20260512/deerflow-backend-offline-image-20260505.tar
+offline-backend-20260512/extensions_config.json
+offline-backend-20260512/skills/
+offline-backend-20260512/wheelhouse/
+offline-backend-20260512/backend/.deer-flow/data/*.db
+offline-backend-20260512/backend/.deer-flow/data/*.db-*
+offline-backend-20260512/backend/.deer-flow/*
+offline-backend-20260512/backend/.env
+offline-backend-20260512/backend/.env.local
+offline-backend-20260512/start_backend.sh
+offline-backend-20260512/stop_backend.sh
+offline-backend-20260512/check_backend.sh
+
+hermes-agent-main/
+
+
+skills/*
+
+offline-backend-20260512/backend/.deer-flow/*
+
+# Per-user skill_evolution overrides — runtime data written by the API
+offline-backend-20260512/backend/skill_evolution.json
+skills/
+hermes-agent-main/
+.gitignore
+offline-backend-20260512/backend/tests/*
+# Keep the openGauss URL/dialect/schema compatibility regression coverage versioned.
+!offline-backend-20260512/backend/tests/test_gauss_database.py
+!offline-backend-20260512/backend/tests/test_llmwiki_weknora.py
+!offline-backend-20260512/backend/tests/test_llmwiki_knowledge_management.py
+!offline-backend-20260512/backend/tests/test_health_auth_bypass.py
+# Keep formal Word OOXML and embedded-font regression coverage versioned.
+!offline-backend-20260512/backend/tests/test_word_export.py
+# Keep generated-file library aggregation and filtering regression coverage versioned.
+!offline-backend-20260512/backend/tests/test_artifact_library.py
+# Keep the model configuration regression coverage versioned even though
+# this deployment repository otherwise ignores newly added backend tests.
+!offline-backend-20260512/backend/tests/test_model_management.py
+# Keep the forced-research agent runtime gate and sandbox-policy regressions versioned.
+!offline-backend-20260512/backend/tests/test_forced_research_agent.py
+# Keep fixed-question exact-match and streaming regression coverage versioned.
+!offline-backend-20260512/backend/tests/test_fixed_questions.py
+# Keep configurable es_query system-prompt routing regression coverage versioned.
+!offline-backend-20260512/backend/tests/test_es_query_routing_prompt.py
+# Keep the local LLMWiki vector-index and external API regression coverage versioned.
+!offline-backend-20260512/backend/tests/test_llmwiki_local_index_repository.py
+!offline-backend-20260512/backend/tests/test_llmwiki_vector_search.py
+!offline-backend-20260512/backend/tests/test_llmwiki_wiki_sync.py
+!offline-backend-20260512/backend/tests/test_external_llmwiki_search.py
+!offline-backend-20260512/backend/tests/test_llmwiki_wiki_retrieval.py
+!offline-backend-20260512/backend/tests/test_public_llmwiki_embed.py
+# Deep-research reliability regression coverage.
+!offline-backend-20260512/backend/tests/test_deep_research_persistence.py
+!offline-backend-20260512/backend/tests/test_deep_research_event_sink.py
+!offline-backend-20260512/backend/tests/test_deep_research_executor_reliability.py
+!offline-backend-20260512/backend/tests/test_deep_research_start_job.py
+!offline-backend-20260512/backend/tests/test_summarization_human_message.py
+!offline-backend-20260512/backend/tests/test_compact_menu_overrides.py
+# Independent enterprise-research persistence, HTTP and race regressions.
+!offline-backend-20260512/backend/tests/test_enterprise_research_tasks.py
+# Position-roundtable per-turn write/present cap (weak-model duplicate artifacts).
+!offline-backend-20260512/backend/tests/test_position_artifact_cap.py
+# Roundtable coordinator/seat capability isolation regressions.
+!offline-backend-20260512/backend/tests/test_roundtable_role_boundary.py
+# Workflow Studio: schemas / definitions / engine / run-lifecycle regressions
+# plus the concurrency remediation suite (atomic resume, draft CAS, publish
+# counter, event seq) — required in CI per docs/WORKFLOW_STUDIO_BACKEND_REMEDIATION_ZH.md P1-06.
+!offline-backend-20260512/backend/tests/test_workflow_schemas.py
+!offline-backend-20260512/backend/tests/test_workflow_definitions.py
+!offline-backend-20260512/backend/tests/test_workflow_engine.py
+!offline-backend-20260512/backend/tests/test_workflow_runs.py
+!offline-backend-20260512/backend/tests/test_workflow_concurrency.py
+!offline-backend-20260512/backend/tests/test_workflow_schema_rebuild.py
+# Writing-mode first-report pipeline enforcement regressions.
+!offline-backend-20260512/backend/tests/test_writing_mode_report_middleware.py
+!offline-backend-20260512/backend/tests/test_writing_rewrite_prompt.py
+# LLM call timeout retry classification regression coverage.
+!offline-backend-20260512/backend/tests/test_llm_error_handling_middleware.py
+# Writing-mode pipeline conversation-material regressions.
+!offline-backend-20260512/backend/tests/test_deep_research_report_tool.py
+# Report collaboration (RC-BE-000–018) regression coverage.
+!offline-backend-20260512/backend/tests/test_report_collaboration_*.py
+!offline-backend-20260512/backend/tests/test_ensure_orm_columns_sync.py
+deploy
+.history
+
+# Cloned third-party repos (local only)
+git-clone/
+
+# Repo-root offline dependency pack (local only)
+/dist/
+
+# Frontend offline pack working directories
+frontend-web/.offline-pack-build/
+frontend-web/.offline-extract-test/
+frontend-web-node_modules-offline-*.tar.gz
+
+# Local dependency caches and backend scratch archives
+/.pnpm-store/
+/offline-backend-20260512/backend/backend*.zip
+
+# Local offline bundles and scratch tooling
+/*.tar.gz
+/offline-backend-20260512/*.zip
+/tmp/
+
+.codegraph
+codegraph.mdc
+
+docs/aidev-docs
+docs/backups
+# Frontend skills components must be tracked (excluded from the broad skills/ ignore above)
+!frontend-web/src/components/workspace/skills/
+
+# uv lockfile (not tracked per project preference)
+offline-backend-20260512/backend/uv.lock
+
+!offline-backend-20260512/backend/tests/test_wiki_package_roundtrip.py
+
+# Local-only scratch directory (not committed)
+0622/
+
+# Python bytecode cache
+__pycache__/
+
+# misc
+.DS_Store

AGENTS.md

@@ -96,7 +96,7 @@ The frontend (`frontend-web/`) is a React 19 + Vite SPA:
 - **智能体使用引导 (Agent onboarding tour, 三段式)**: 智能体「浏览管理 → 创建与配置 → 发起问答」全流程分步引导，组件 `components/workspace/agents/agent-guide.tsx`（同一 TDesign `Guide` 封装模式，与简洁模式引导共享记忆语义：**仅显式点「完成/跳过」才写 localStorage** `deerflow.agent-guide.<stage>.finished`，中途刷新/离开不记忆）。**自动演示仅 create / chat 两段**（进智能体列表页不弹引导，点「创建智能体」进入新建页时才自动演示；list 段的 `AgentGuide` 传 `autoStart={false}`，只能由头部「使用引导」按钮手动触发）。三段各自挂载、各自记忆：① `list`（`agent-gallery.tsx`，/page/workspace/agents）——分区 Tabs、搜索排序、智能体卡片操作、右上「创建智能体」按钮；**完成/跳过就地结束（不跳转创建页，与 create 段无联动）**，最后一步仅提示进入创建页后会有配置引导；② `create`（`NewAgentPage.tsx`，/page/workspace/agents/new）——左侧对话式创建输入区、基本信息、运行模型+入库模板名、知识空间、SOUL.md、技能穿梭框、保存按钮（9 步）；③ `chat`（`AgentChatPage.tsx`，agents/:id/chats）——左栏切换与历史、常用问题 chips、输入框（落地态/会话态两个挂点同一 `chat-input` 标记）、右侧详情面板；`eligible` 限定普通智能体对话页（`showAgentList && !isEmbedded && !taskCtxOn`，即排除通用问答原地嵌入与任务深链工作区），三段共用 `isAgentGuideEnvironment()` 排除 embed/hideChrome/无界环境。高亮锚点 `data-agent-guide`：列表页 `tabs`/`search`/`agent-card`（卡片+列表两种布局的根元素）/`new-agent-btn`，创建页 `create-chat-input`/`create-basic`/`create-model`/`create-knowledge`/`create-soul`/`create-skills`/`create-save-btn`，对话页 `chat-sidebar`/`chat-questions`/`chat-input`/`chat-detail`；同样**启动时按目标存在性过滤步骤**（新用户无卡片/无预设问题时自动剔除）。手动重放：三个页面头部各有「使用引导」按钮（列表页创建按钮左侧、创建页 header 右侧、对话页 header），广播 `agent-guide:start:<stage>` 窗口事件强制重放（忽略已看过记忆）。
 - **敏感词管理 (Sensitive Word Management)**: configures the **展示层脱敏映射** (原始词 → 替换词). **Gated on the username `lqq`** (not the admin role) — `getAccountDisplayName().toLowerCase() === "lqq"` (username = email `@` prefix, else `user_id`, same as the profile page); the dropdown entry, the route guard, and the page-level guard all use this check. `src/lib/desensitize.ts` ships a frozen seed map `SEED_DESENSITIZE_MAP` (the deployment seed) plus a live mutable `DESENSITIZE_MAP` (initialized from the seed so `desensitize()` works before any fetch). `DesensitizeWordsLoader` (mounted in `App.tsx` `AppProviders`) fetches `/api/sensitive-words` on startup and, if non-empty, calls `replaceDesensitizeMap(...)` to make the backend config the source of truth (empty/failed → keep seed). Page `SensitiveWordsPage.tsx` (route `/page/strategy/admin/sensitive-words`, opened from the 「敏感词管理」 item in the **bottom-left settings dropdown** of `components/page-sidebar.tsx`, present in **both** full + collapsed sidebar modes) lists/adds/edits/启停/deletes words and saves the **whole set atomically** (replace-all); a 「导入种子词」 button merges any missing `SEED_DESENSITIZE_MAP` entries into the editor. Client + hooks in `strategy-components/api/sensitive-words.ts` (`wordsToMap` folds enabled rows into the runtime map). Backend: `/api/sensitive-words` (global/shared; open read, write allowed for admin **or** the `lqq` account), mirroring 菜单管理's store shape. **侧边栏 + 导航栏 menu labels are desensitized** at the data source: `useSidebarMenuItems`/`useActiveThirdLevel` (`core/page-layout/use-sidebar-menu.ts`) map every node's `label` through a desensitizer built **directly from the `/api/sensitive-words` query** (`makeDesensitizer(wordsToMap(words))`, falling back to the global seed `desensitize` while empty) — this sidesteps the `replaceDesensitizeMap` effect-timing race. Only `label` is rewritten; `path`/`id`/`icon` are untouched so routing/highlight are unaffected. All nav surfaces (`page-sidebar.tsx`, `page-sidebar-dark.tsx`, `page-sidebar-v1.tsx`, `strategy-components/Header.tsx`, `PageRoutes.tsx`) consume this hook and are covered; `MenuManagementPage` uses the raw tree so admins still edit real labels.
 - **技能地址管理 (Skill Address Management, admin-only)**: 扫描运行中的技能（`skills/{public,custom}` 下所有 `.md`/`.py`）里的 **http(s) URL 与 IP/IP:端口**，按唯一地址聚合全部出现位置（跨 .md/.py），支持**批量替换**（如把 UAT 地址一次性改成生产）。主体抽成可复用组件 `SkillAddressesPanel`（`pages/SkillAddressesPage.tsx` 导出，默认导出 `SkillAddressesPage` 仍保留旧路由 `/page/strategy/admin/skill-addresses` 作直链回退），**作为「技能地址管理」标签页嵌入管理员「技能管理」页 `pages/AdminSkillsPage.tsx`**（该页加 Tabs：「技能列表」+「技能地址管理」；入口在「设置和更多」下拉的「技能管理」`workspace-nav-menu.tsx` → `/page/workspace/admin/skills`，不再单列于 `components/page-sidebar.tsx` 左下角设置下拉）：打开自动扫描并带**进度条**（扫描逐技能 / 应用逐文件 X/N，应对 150+ 技能），URL/IP 类型过滤 + 搜索 + 「片段批量套用」（如所有 `uat.4.cn` → `prod.4.cn`）、每地址「替换为」输入框（可展开看 `file:line` 片段）、内联**预览**、**应用**、**多关键词检索**（空格分隔全部命中），以及**修改记录**面板——每次应用生成一条记录（显示时间/改了哪些地址 from→to/文件数/处数），可**针对单次记录回退**（二次确认）。client + NDJSON 流式解析在 `strategy-components/api/skill-addresses.ts`。后端 `/api/skill-addresses`（**仅管理员**）：所有文件 IO 走 `asyncio.to_thread` 不阻塞事件循环；`GET /scan`(+`/scan/stream`)、`POST /preview`、`POST /apply`(+`/apply/stream`)、`GET /backups`、`POST /rollback/{id}`；**子串安全替换**（基于 token span，不会误伤子串）、`expected_total` 乐观锁（409）、写前自动磁盘备份到 `{skills_root}/.address-edits/`（同时写一份 `{id}.manifest.json` 记录本次替换明细，供修改记录展示）、应用后刷新技能系统提示缓存。纯引擎 `deerflow/skills/address_scan.py`，router `app/gateway/routers/skill_addresses.py`，tests `tests/test_skill_addresses.py`。范围仅运行中技能，不动 `deploy/minimal/skills/`。
-- **按钮管理 (Task Button Management, admin-only)**: configures the **可配置跳转按钮** rendered in the left task panel of the task deep-link workspaces (rwfx / 3qfx / xdfx, see `TaskFxWorkspace`). Each button = `{business, label, linkType, target, appendTaskId, idParam, idKind, enabled, sortOrder}`; `linkType` is `url` (外链 `window.open`, optional append of `?{idParam}={id}` — **`idParam` is the editable query-param name**, e.g. `id`/`task_id`), `business` (内部切换到另一业务, reuses `resolveTaskDeeplinkTarget`, always task id), or `purpose` (弹出当前任务的「任务目的」表格 via the shared `TaskPurposeDialog`, no target — configurable for any page incl. xdfx). **`host`（宿主路由）** asks the embedding parent to navigate: `runTaskButton` sends `window.parent.postMessage({ type:'HOST_NAVIGATE', path, queryParams:{ [idParam]: Number(taskId) } }, hostOrigin)` — `target` is the host route, `appendTaskId`/`idParam` become numeric `queryParams` (task id only), origin from `runtime-config.js` `VITE_HOST_NAVIGATE_ORIGIN` (else ancestorOrigins/referrer; never `*`). **`idKind` (`task`/`action`) only matters for xdfx**: its deep-link `?taskId=` is an 行动id and the 任务id is resolved at login (`fetchTaskIdFromAction`) — `TaskContext` now also stashes the raw `actionId`, so a url button can append either (default `task`). Page `TaskButtonsPage.tsx` (route `/page/strategy/admin/task-buttons`, **admin-gated**, entry「按钮管理」in the bottom-left settings dropdown of `components/page-sidebar.tsx`, both full + collapsed modes) has per-business tabs, edits name/type/target/启停/排序/增删, saves the **whole set atomically** (replace-all), plus 「导入默认」 to merge code-seed defaults. **Seed/fallback** (like 敏感词管理): only when the whole table is empty (fresh deploy, never saved) do the workspace + page fall back to `defaultTaskButtons(cfg)` — rwfx's url buttons (编辑/下一步/3q详情) built from the **live** `/api/public/config/task-deeplink` plus cross-business switch buttons for **all three** businesses (each switches to the other two — rwfx→三情/行动, 3qfx→任务/行动, xdfx→任务/三情). Once saved, the store is the sole source of truth. The old hardcoded jump buttons were migrated out: `TaskFxWorkspace` no longer hardcodes 切换到三情分析, `TaskActionsPanel` keeps only the non-jump 清除记录, and `TaskPurposePanel` dropped its 3q详情 button (now configurable) **and defaults the 目的树 table to collapsed** (chevron toggle). Helpers `core/auth/task-buttons.ts` (`TaskButton`, `defaultTaskButtons`, `resolveTaskButtons`, `buttonsForBusiness`, `runTaskButton`), client+hooks `strategy-components/api/task-buttons.ts`. Backend: `/api/task-buttons` (global/shared; open read, admin replace-all), store `deerflow.persistence.task_buttons` (table `task_buttons`, auto-created by `create_all`), wired `app.state.task_button_store`. Tests: `tests/test_task_buttons.py`.
+- **按钮管理 (Task Button Management, admin-only)**: configures the **可配置跳转按钮** rendered in the left task panel of the task deep-link workspaces (rwfx / 3qfx / xdfx, see `TaskFxWorkspace`). Each button = `{business, label, linkType, target, appendTaskId, idParam, idKind, enabled, sortOrder}`; `linkType` is `url` (外链 `window.open`, optional append of `?{idParam}={id}` — **`idParam` is the editable query-param name**, e.g. `id`/`task_id`), `business` (内部切换到另一业务, reuses `resolveTaskDeeplinkTarget`, always task id), or `purpose` (弹出当前任务的「任务目的」表格 via the shared `TaskPurposeDialog`, no target — configurable for any page incl. xdfx). **`host`（宿主路由）** asks the embedding parent to navigate: `runTaskButton` sends `window.parent.postMessage({ type:'HOST_NAVIGATE', path, queryParams:{ [idParam]: Number(taskId) } }, hostOrigin)` — `target` is the host route, `appendTaskId`/`idParam` become numeric `queryParams` (task id, or action id when `idKind:"action"` — same selector as url buttons, xdfx only), origin from `runtime-config.js` `VITE_HOST_NAVIGATE_ORIGIN` (else ancestorOrigins/referrer; never `*`). **`idKind` (`task`/`action`) only matters for xdfx**: its deep-link `?taskId=` is an 行动id and the 任务id is resolved at login (`fetchTaskIdFromAction`) — `TaskContext` now also stashes the raw `actionId`, so a url **or host** button can append either (default `task`). Page `TaskButtonsPage.tsx` (route `/page/strategy/admin/task-buttons`, **admin-gated**, entry「按钮管理」in the bottom-left settings dropdown of `components/page-sidebar.tsx`, both full + collapsed modes) has per-business tabs, edits name/type/target/启停/排序/增删, saves the **whole set atomically** (replace-all), plus 「导入默认」 to merge code-seed defaults. **Seed/fallback** (like 敏感词管理): only when the whole table is empty (fresh deploy, never saved) do the workspace + page fall back to `defaultTaskButtons(cfg)` — rwfx's url buttons (编辑/下一步/3q详情) built from the **live** `/api/public/config/task-deeplink` plus cross-business switch buttons for **all three** businesses (each switches to the other two — rwfx→三情/行动, 3qfx→任务/行动, xdfx→任务/三情). Once saved, the store is the sole source of truth. The old hardcoded jump buttons were migrated out: `TaskFxWorkspace` no longer hardcodes 切换到三情分析, `TaskActionsPanel` keeps only the non-jump 清除记录, and `TaskPurposePanel` dropped its 3q详情 button (now configurable) **and defaults the 目的树 table to collapsed** (chevron toggle). Helpers `core/auth/task-buttons.ts` (`TaskButton`, `defaultTaskButtons`, `resolveTaskButtons`, `buttonsForBusiness`, `runTaskButton`), client+hooks `strategy-components/api/task-buttons.ts`. Backend: `/api/task-buttons` (global/shared; open read, admin replace-all), store `deerflow.persistence.task_buttons` (table `task_buttons`, auto-created by `create_all`), wired `app.state.task_button_store`. Tests: `tests/test_task_buttons.py`.
 - **State**: TanStack Query for server state, context providers (`SubtasksProvider`, `ArtifactsProvider`, `PromptInputProvider`) scoped to `WorkspaceRoutes`
 - **LLM streaming**: `@langchain/langgraph-sdk` streams messages from the backend via SSE
 - **UI**: Radix UI primitives + TailwindCSS v4 + shadcn-style components in `src/components/ui/`

README.md

@@ -1,3 +0,0 @@
-# deerflow-code
-
-deerflow-code

docs/agentscope-多智能体报告协作工作台-交互重设计方案.md

@@ -0,0 +1,186 @@
+# 报告协作工作台 — 交互重设计方案（v2）
+
+> 日期：2026-09-07
+> 状态：**待确认**（确认后按 §7 分期实施）
+> 取代：现行「规则澄清 + 三张模板方案卡 + 画布为主」的交互
+> 原则：回到最初设计 —— **协调者智能体全程与用户对话的主线**
+
+---
+
+## 1. 设计目标与总体形态
+
+用户视角只有一条主线：**和一个规划师智能体对话 → 它摸清意图并规划工作流 → 确认后执行、过程在流程图上可见 → 随时插话干预、指定智能体重跑 → 最终成稿、继续对话式修改**。
+
+```
+┌─────────────────────────────────────────────────────────┐
+│  阶段1 计划模式        阶段2/3 执行+干预      阶段4 成稿  │
+│  ┌───────────┐        ┌───────────┐        ┌──────────┐ │
+│  │ 规划师对话 │  确认   │ 聊天时间线 │  审稿过 │ 对话改稿 │ │
+│  │ 需求摘要卡 │ ─────→ │ +成员Tab  │ ─────→ │ 版本/diff│ │
+│  │ 工作流卡   │  冻结   │ +流程图动画│  报告卡 │ 上游重跑 │ │
+│  └───────────┘        └───────────┘        └──────────┘ │
+│        ↑ 回到计划模式（replan）随时可触发 ↑                │
+└─────────────────────────────────────────────────────────┘
+```
+
+与现行的核心差异：
+
+| 维度 | 现行（做偏的） | v2 |
+|------|----------------|-----|
+| 需求澄清 | 规则模板问句秒回 | 规划师智能体（LLM）领域化提问 |
+| 工作流来源 | 3 张确定性模板 | LLM 按意图生成自定义 DAG，对话迭代修改 |
+| 执行展示 | 画布为主、聊天为辅 | **聊天时间线为主**（成员 Tab 切换保留），**执行状态在流程图上动画呈现** |
+| 干预 | 命令确认卡体系 | 输入框常开 → 协调者智能体分类；支持**指定智能体重跑** |
+| 报告修改 | 改写候选（纯文字层） | 意图分流：润色走改写；**事实性问题触发上游节点重跑** |
+
+---
+
+## 2. 阶段 1：计划模式 — 需求梳理与工作流规划
+
+### 2.1 计划模式状态
+
+- 用户发出第一条需求消息起进入**计划模式**，Header 显示徽章「📋 计划模式」，输入框提示语切换为「梳理需求中，你可以随时补充或纠正…」。
+- 计划模式在**工作流确认冻结**时结束；执行期大改可通过 replan 回到计划模式（见 §5.4）。
+
+### 2.2 规划师对话（LLM，替换现行规则路由）
+
+- 模型：`report_collaboration.planner_model`（回退 `coordinator_model` → 默认模型，复用现有 model_resolver 冻结机制）。
+- 提问是领域化的：写市场分析就问「关注整车还是供应链 / 要不要竞品对比 / 给谁看 / 篇幅」，而非固定模板问句。
+- 澄清轮次上限沿用 `max_clarification_rounds`；用户随时可以说「直接开始」按默认假设推进。
+
+### 2.3 需求摘要卡
+
+信息足够后，规划师在对话中输出一张**结构化需求摘要卡**（业务卡插槽渲染）：主题 / 目的 / 读者 / 角度清单 / 篇幅与重点 / 默认假设。用户逐项看，可在对话里继续改（「读者改成投资人」→ 卡片更新出新版）。
+
+### 2.4 工作流生成与迭代
+
+- 用户确认摘要后，规划师生成**一份**工作流方案（不是三选一）：以工作流卡（成员构成、各节点使命、预计时长/成本）+ 右侧流程图预览呈现。
+- 生成约束（保证可执行）：**节点类型受词表约束** — `research / verify / analyze / write / review`；自由的是结构（几条线、并行还是串行、核验深度）与每个节点的使命/角度/工具。
+- 对话式修改：「把价格策略单独拆一条线」「加一个政策风险分析」→ 规划师修订后出新版工作流卡（带变更说明），再确认。
+- 「开始执行」→ 冻结需求 + 工作流（含模型/预算快照，沿用现有 freeze 机制）。
+
+---
+
+## 3. 阶段 2：执行 — 过程在流程图上可见
+
+### 3.1 流程图动画规格（执行状态的可视化主载体）
+
+节点状态 → 样式（`ReportFlowCanvas` / `ReportFlowNode`，状态源 `selectNodeStatusMap`，事件源 `node.status.changed` SSE）：
+
+| 状态 | 节点样式 | 连入边 |
+|------|----------|--------|
+| planned / queued | 灰色常态 | 静态 |
+| **running** | 主色描边 + **呼吸脉冲动画**（ring）+ 角标 spinner | **流动动画**（@xyflow 内置 `animated: true` 的 dash 流动） |
+| **completed** | 绿色 + ✓ 角标，常驻 | 恢复静态实线 |
+| failed | 红色 + 错误角标 | 红色虚线 |
+| awaiting_input | 橙色 + ？角标 | 橙色静态 |
+| interrupted / superseded | 半透明降级 | 灰 |
+
+- 语义：**边流动 = 数据正在流入该节点**（该节点执行中）；节点完成即停动画换完成样式。
+- 实现成本：@xyflow 原生支持 animated 边与自定义节点 className，纯 CSS 补呼吸动画即可，无需新依赖。
+
+### 3.2 聊天区（保留现有 Tab 切换）
+
+- 并行成员沿用 **AgentStageSelector Tab 切换**（维持现状，不改交错时间线）。
+- 每个 Tab 内是该成员的完整流式输出：思考、工具调用卡、检索过程、正文 —— 与普通智能体问答同观感（message-adapter 已具备映射能力）。
+- 协调者播报（「研究线 A 完成，核心发现 X，开始交叉核验」）进入总时间线 Tab。
+
+---
+
+## 4. 阶段 3：干预 — 对话即干预 + 指定重跑
+
+### 4.1 输入框常开，协调者智能体分流
+
+执行期输入框永远可用，所有输入先到协调者（LLM 分类，规则高精度直判兜底——复用 RC-BE-014 IntentRouter 骨架，把「规则优先」反转为「LLM 优先 + 规则快路径」）：
+
+| 用户意图 | 协调者行为 | 生效时机 |
+|----------|-----------|----------|
+| 提问（「查到什么了？」） | 直接回答 | 立即，不打断 |
+| 修正方向（「重点看下沉市场」） | 小改：在安全点注入 overlay，并回复说明 | 下个安全点 |
+| **指定重跑**（见 §4.2） | 出确认卡 → 重开节点 attempt | 确认后 |
+| 大改（砍线/加线） | 提议 replan → 回计划模式 | 用户确认 |
+| 停止 | 直接取消 | 立即 |
+
+### 4.2 指定智能体重跑（新增核心交互）
+
+**场景**：用户发现某节点传给下游的信息不对（如研究线 B 的结论有误）。
+
+两个入口，同一后端动作：
+
+1. **对话**：「研究线 B 的结果不对，重新查，重点看价格战」→ 协调者解析出 `node_id` + 新指令；
+2. **画布**：点节点 → 抽屉（NodeInspectorDrawer）加「↻ 重新执行」按钮。
+
+**级联规则**（复用并补全 ledger 的 targeted rework）：
+
+```
+重跑节点 N（attempt+1，带用户新指令作为 mission overlay）
+  → N 完成 → 所有直接/间接依赖 N 产物的下游节点标记 superseded
+  → 按 DAG 依赖序逐波重跑（边动画重新流动，用户全程可见）
+  → 受影响章节重新生成
+```
+
+- 确认卡必须展示**影响面**：「将级联重跑 核验B / 分析 / 写作 共 3 个节点，预计 +8 分钟」。
+- 成本保护：级联重跑计入 run 预算（现有 budget 机制），预算不足时明确报因。
+
+---
+
+## 5. 阶段 4：报告成稿与对话式修改
+
+### 5.1 成稿
+
+审稿通过 → 报告卡出现在时间线（点开 ReportPreviewDialog 全文），协调者播报总结（各线核心发现 → 结论如何形成）。
+
+### 5.2 修改意图分流（协调者 LLM 判定）
+
+| 用户说 | 判定 | 动作 |
+|--------|------|------|
+| 「把第二章结论写得更笃定」 | 文字润色 | 改写候选（现有 rewrite API）→ 对比预览 → apply → 新版本 |
+| 「这个数据不对 / 重新收集一下」 | **事实性问题** | 定位来源节点 → **触发该节点重跑（可带新角度）→ 级联 §4.2 → 受影响章节重生成 → 新版本** |
+| 「再从供应链角度补一节」 | 结构性新增 | replan：回计划模式改工作流（只增不删已完成部分） |
+| 「恢复到上一版」 | 版本回退 | 现有 restore API |
+
+- 版本链：每次 apply / restore / 级联重跑产出均产生新 head，历史完整保留（现有机制）。
+- 修改对话里协调者必须先**复述理解**（「你是想让研究线 B 换个角度重查，然后重写第三章，对吗？」）再执行，避免误触发昂贵的级联。
+
+---
+
+## 6. 工程改造清单
+
+### 6.1 保留（本次联调已验证）
+
+- 持久化 13 张表 + durable SSE + `after_seq` 断线续传；幂等键 / 乐观锁 / 预算框架。
+- MessageList 复用链路（message-adapter → LangGraph Message → 业务卡插槽）。
+- AgentStageSelector Tab、commands 安全点骨架（RC-BE-014）、报告版本三件套（RC-BE-015）、前端 hooks/reducer/selectors 主体。
+- 「清空历史」「生成候选方案」等本次新增入口。
+
+### 6.2 后端重做 / 新建
+
+| 项 | 说明 |
+|----|------|
+| 规划师智能体 | LLM 多轮澄清（替换 `requirement/intent_router.py` 规则路由）+ 需求摘要结构化 + **工作流 LLM 生成**（受 §2.4 词表约束，替换 `planning/controller.py` 三模板）+ 对话式工作流修订 |
+| 协调者智能体升级 | 干预分类 LLM 优先；修改意图分流（§5.2）；执行前复述确认 |
+| 指定重跑 + 级联 | ledger targeted rework 补全级联 superseded/重跑波次；确认卡带影响面计算 |
+| AgentScope 适配器修复 | `count_tokens` 协议缺失（上下文压缩必崩）；attempt 耗尽终态化（防僵尸 run） |
+| 事件补齐 | 边流动所需的节点状态事件已有；补「级联影响预告」事件 |
+
+### 6.3 前端重做 / 新建
+
+| 项 | 说明 |
+|----|------|
+| 计划模式 UI | Header 徽章、composer 提示切换、需求摘要卡、工作流卡（替换现三张模板卡） |
+| 流程图动画 | running 节点呼吸动画 + 入边 `animated`、completed ✓ 样式、级联重跑时边动画重流（§3.1） |
+| 节点抽屉 | 加「重新执行」按钮 → 确认卡 |
+| 干预交互 | 执行期输入框常开；意图确认卡（含影响面）；修改对话复述确认 |
+| 布局 | 聊天为主、画布为辅的默认比例微调（不动整体两栏结构） |
+
+---
+
+## 7. 实施分期
+
+| 期 | 内容 | 验收 |
+|----|------|------|
+| **P0 地基** | AgentScope `count_tokens` 修复；attempt 耗尽终态化；级联重跑的 ledger 补全 | 执行阶段不崩、无僵尸 run、单节点重跑级联正确（后端测试） |
+| **P1 计划模式 + 执行可视化** | 规划师 LLM 对话、需求摘要卡、工作流 LLM 生成与迭代、计划模式 UI、流程图动画 | 从新会话到确认工作流全程对话完成；执行时动画正确随状态流转 |
+| **P2 干预 + 成稿修改** | 协调者 LLM 分类、指定重跑（两入口 + 影响面确认卡）、报告修改意图分流、上游级联重跑 | 四类干预各自走通；「数据不对→重跑→新章节→新版本」全链路 |
+
+每期结束跑 RC 全量测试（当前基线 233 通过）+ 前端 26 单测，不引入对其他模块的影响。

frontend-web/src/core/auth/task-buttons.ts

@@ -109,8 +109,8 @@ export interface TaskButton {
   /** 追加的查询参数名（如 ``id`` / ``task_id``）；空则用 ``id``。 */
   idParam: string;
   /**
-   * ``url`` 追加 id 时取任务id 还是行动id。仅 xdfx（行动分析）二者不同：深链原始 ``?taskId=``
+   * ``url`` / ``host`` 追加 id 时取任务id 还是行动id。仅 xdfx（行动分析）二者不同：深链原始
-   * 实为行动id，任务id 由 action 接口解析。其它业务两者相同。
+   * ``?taskId=`` 实为行动id，任务id 由 action 接口解析。其它业务两者相同。
    */
   idKind: TaskButtonIdKind;
   /** ``url`` 外链打开方式：``blank``=新窗口（默认）/ ``self``=当前窗口。仅 ``url`` 类型有意义。 */
@@ -327,7 +327,8 @@ export interface RunTaskButtonCtx {
  * 执行一个跳转按钮：
  *   - ``url``：拼地址（可选追加 ``?id=``，取任务id 或行动id 视 ``idKind``）新标签打开；地址为空提示未配置。
  *   - ``business``：复用深链解析（会按目标业务重绑任务上下文）后 ``navigate``（始终用任务id）。
- *   - ``host``：向 ``window.parent`` 发 ``HOST_NAVIGATE``，由宿主系统跳转（仅 iframe 嵌入时生效）。
+ *   - ``host``：向 ``window.parent`` 发 ``HOST_NAVIGATE``，由宿主系统跳转（仅 iframe 嵌入时生效；
+ *     追加 id 时同样按 ``idKind`` 取任务id / 行动id）。
  */
 export async function runTaskButton(btn: TaskButton, ctx: RunTaskButtonCtx): Promise<void> {
   const fail = (m: string) => ctx.onError?.(m);
@@ -355,9 +356,11 @@ export async function runTaskButton(btn: TaskButton, ctx: RunTaskButtonCtx): Pro
     }
     const queryParams: Record<string, number> = {};
     if (btn.appendTaskId) {
-      const id = parseHostTaskId(ctx.taskId);
+      // 行动分析（xdfx）可选发行动id；与 url 外链同规则：行动id 缺失时回退任务id。
+      const raw = btn.idKind === "action" ? ctx.actionId || ctx.taskId : ctx.taskId;
+      const id = parseHostTaskId(raw);
       if (id === null) {
-        fail("任务id不是数字，无法跳转宿主系统");
+        fail(`${btn.idKind === "action" ? "行动id" : "任务id"}不是数字，无法跳转宿主系统`);
         return;
       }
       queryParams[(btn.idParam || "id").trim() || "id"] = id;

frontend-web/src/core/skills/api.ts

@@ -0,0 +1,637 @@
+import { getBackendBaseURL } from "@/core/config";
+import { apiFetch } from "@/core/api";
+
+import type { Skill, SkillDisplayConfig, SkillState } from "./type";
+
+export interface ListSkillsParams {
+  search?: string;
+  tagIds?: string[];
+  sort?: "time_desc" | "time_asc" | "activity_desc" | "activity_asc" | "usage_desc" | "usage_asc" | "name_asc" | "name_desc";
+  /** Keep only skills filed under this publish square (发布广场). */
+  square?: string;
+  /** ``"mine"`` restricts to the skills the caller's 岗位 (position) makes visible. */
+  scope?: "mine";
+}
+
+export async function loadSkills(params: ListSkillsParams = {}) {
+  const qs = new URLSearchParams();
+  if (params.search) qs.set("search", params.search);
+  for (const id of params.tagIds ?? []) qs.append("tag_ids", id);
+  if (params.sort) qs.set("sort", params.sort);
+  if (params.square) qs.set("square", params.square);
+  if (params.scope) qs.set("scope", params.scope);
+  const query = qs.toString();
+  const skills = await apiFetch(
+    `${getBackendBaseURL()}/api/skills${query ? `?${query}` : ""}`,
+  );
+  const json = await skills.json();
+  return json.skills as Skill[];
+}
+
+export async function loadAllSkillsAdmin() {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills?admin=true`,
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to load skills");
+  }
+  const json = await response.json();
+  return json.skills as Skill[];
+}
+
+export async function getSkill(skillName: string): Promise<Skill> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}`,
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to load skill");
+  }
+  return response.json() as Promise<Skill>;
+}
+
+export async function enableSkill(skillName: string, enabled: boolean) {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${skillName}`,
+    {
+      method: "PUT",
+      headers: {
+        "Content-Type": "application/json",
+      },
+      body: JSON.stringify({
+        enabled,
+      }),
+    },
+  );
+  return response.json();
+}
+
+export async function setSkillAlwaysOn(
+  skillName: string,
+  alwaysOn: boolean,
+): Promise<Skill> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/always-on`,
+    {
+      method: "PUT",
+      headers: {
+        "Content-Type": "application/json",
+      },
+      body: JSON.stringify({ always_on: alwaysOn }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to update skill 常驻 flag");
+  }
+  return response.json() as Promise<Skill>;
+}
+
+export async function setSkillDisplay(
+  skillName: string,
+  display: SkillDisplayConfig | null,
+  sample?: Record<string, unknown> | null,
+): Promise<Skill> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/display`,
+    {
+      method: "PUT",
+      headers: {
+        "Content-Type": "application/json",
+      },
+      body: JSON.stringify({ display, sample }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to update skill display adapter");
+  }
+  return response.json() as Promise<Skill>;
+}
+
+export interface SkillDisplaySampleResponse {
+  skill_name: string;
+  sample: Record<string, unknown>;
+}
+
+export interface PersistedSkillDisplaySampleResponse {
+  skill_name: string;
+  sample: Record<string, unknown> | null;
+}
+
+export async function getPersistedSkillDisplaySample(
+  skillName: string,
+): Promise<PersistedSkillDisplaySampleResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/display-sample`,
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to get persisted skill display sample");
+  }
+  return response.json() as Promise<PersistedSkillDisplaySampleResponse>;
+}
+
+export async function getSkillDisplaySample(
+  skillName: string,
+): Promise<SkillDisplaySampleResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/display-sample`,
+    { method: "POST" },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to get skill display sample");
+  }
+  return response.json() as Promise<SkillDisplaySampleResponse>;
+}
+
+export interface InstallSkillRequest {
+  thread_id: string;
+  path: string;
+  skill_name?: string;
+}
+
+export interface InstallSkillResponse {
+  success: boolean;
+  skill_name: string;
+  message: string;
+  overwritten?: boolean;
+}
+
+export interface PreviewSkillInstallRequest {
+  thread_id: string;
+  path: string;
+}
+
+export interface PreviewSkillInstallResponse {
+  ok: boolean;
+  skill_name: string;
+  description: string;
+  license: string | null;
+  allowed_tools: string[];
+  can_install_as_is: boolean;
+  owned_by_me: boolean;
+  owned_by_other: boolean;
+  builtin_conflict: boolean;
+  owner_username?: string | null;
+  owner_user_id?: string | null;
+  message: string;
+  warnings: string[];
+}
+
+export interface ValidateSkillUploadResponse {
+  ok: boolean;
+  skill_name: string;
+  description: string;
+  license: string | null;
+  allowed_tools: string[];
+  conflict: boolean;
+  conflict_kind: "public" | "custom" | null;
+  can_install_as_is: boolean;
+  can_overwrite: boolean;
+  owned_by_me: boolean;
+  owned_by_other: boolean;
+  builtin_conflict: boolean;
+  owner_username?: string | null;
+  owner_user_id?: string | null;
+  message: string;
+  warnings: string[];
+}
+
+export async function installSkill(
+  request: InstallSkillRequest,
+): Promise<InstallSkillResponse> {
+  const response = await apiFetch(`${getBackendBaseURL()}/api/skills/install`, {
+    method: "POST",
+    headers: {
+      "Content-Type": "application/json",
+    },
+    body: JSON.stringify(request),
+  });
+
+  if (!response.ok) {
+    // Handle HTTP error responses (4xx, 5xx)
+    const errorData = await response.json().catch(() => ({}));
+    const errorMessage =
+      errorData.detail ?? `HTTP ${response.status}: ${response.statusText}`;
+    return {
+      success: false,
+      skill_name: "",
+      message: errorMessage,
+    };
+  }
+
+  return response.json();
+}
+
+export async function previewSkillInstall(
+  request: PreviewSkillInstallRequest,
+): Promise<PreviewSkillInstallResponse> {
+  const response = await apiFetch(`${getBackendBaseURL()}/api/skills/install-preview`, {
+    method: "POST",
+    headers: {
+      "Content-Type": "application/json",
+    },
+    body: JSON.stringify(request),
+  });
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to preview skill install");
+  }
+  return response.json() as Promise<PreviewSkillInstallResponse>;
+}
+
+async function parseDetailedError(response: Response, fallback: string): Promise<Error> {
+  const data = await response.json().catch(() => ({}));
+  const detail = (data as { detail?: unknown }).detail;
+  const message = typeof detail === "string" && detail ? detail : `${fallback} (HTTP ${response.status})`;
+  return new Error(message);
+}
+
+export async function validateSkillUpload(
+  file: File,
+  targetCategory: "public" | "custom" = "custom",
+): Promise<ValidateSkillUploadResponse> {
+  const form = new FormData();
+  form.append("file", file, file.name);
+  const params = new URLSearchParams({ target_category: targetCategory });
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/validate?${params.toString()}`,
+    {
+      method: "POST",
+      body: form,
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Validation failed");
+  }
+  return response.json() as Promise<ValidateSkillUploadResponse>;
+}
+
+export async function installSkillUpload(
+  file: File,
+  options: { overwrite?: boolean; targetCategory?: "public" | "custom" } = {},
+): Promise<InstallSkillResponse> {
+  const form = new FormData();
+  form.append("file", file, file.name);
+  const params = new URLSearchParams();
+  if (options.overwrite) params.set("overwrite", "true");
+  if (options.targetCategory) params.set("target_category", options.targetCategory);
+  const query = params.size > 0 ? `?${params.toString()}` : "";
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/install-upload${query}`,
+    { method: "POST", body: form },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Install failed");
+  }
+  return response.json() as Promise<InstallSkillResponse>;
+}
+
+export async function duplicateSkill(
+  sourceName: string,
+  newName: string,
+): Promise<InstallSkillResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/custom/${encodeURIComponent(sourceName)}/duplicate`,
+    {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ new_name: newName }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to duplicate skill");
+  }
+  return response.json() as Promise<InstallSkillResponse>;
+}
+
+export interface ReconcileCustomSkillsResponse {
+  adopted: string[];
+  already_registered: string[];
+}
+
+export async function reconcileCustomSkills(): Promise<ReconcileCustomSkillsResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/reconcile-custom`,
+    { method: "POST" },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to reconcile skills directory");
+  }
+  return response.json() as Promise<ReconcileCustomSkillsResponse>;
+}
+
+export interface SetSkillPublishedResponse {
+  name: string;
+  owner_user_id: string | null;
+  published: boolean;
+  square_id?: string;
+}
+
+export async function setSkillPublished(
+  skillName: string,
+  published: boolean,
+  squareId?: string,
+): Promise<SetSkillPublishedResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/custom/${encodeURIComponent(skillName)}/published`,
+    {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(
+        squareId != null ? { published, square_id: squareId } : { published },
+      ),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to update published state");
+  }
+  return response.json() as Promise<SetSkillPublishedResponse>;
+}
+
+export async function setSkillState(
+  skillName: string,
+  state: SkillState,
+): Promise<Skill> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/custom/${encodeURIComponent(skillName)}/state`,
+    {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ state }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to update skill state");
+  }
+  return response.json() as Promise<Skill>;
+}
+
+export async function setSkillPinned(
+  skillName: string,
+  pinned: boolean,
+): Promise<Skill> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/custom/${encodeURIComponent(skillName)}/pinned`,
+    {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ pinned }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to update skill pinned state");
+  }
+  return response.json() as Promise<Skill>;
+}
+
+export async function setSkillFeatured(
+  skillName: string,
+  featured: boolean,
+): Promise<Skill> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/custom/${encodeURIComponent(skillName)}/featured`,
+    {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ featured }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to update skill featured state");
+  }
+  return response.json() as Promise<Skill>;
+}
+
+export interface SkillContentResponse {
+  name: string;
+  category: string;
+  content: string;
+}
+
+export async function getSkillContent(
+  skillName: string,
+): Promise<SkillContentResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/content`,
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to load skill content");
+  }
+  return response.json() as Promise<SkillContentResponse>;
+}
+
+export async function deleteCustomSkill(skillName: string): Promise<void> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/custom/${encodeURIComponent(skillName)}`,
+    { method: "DELETE" },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to delete skill");
+  }
+}
+
+/** Delete a built-in skill. The backend permits this only for administrators. */
+export async function deleteBuiltinSkill(skillName: string): Promise<void> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}`,
+    { method: "DELETE" },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to delete built-in skill");
+  }
+}
+
+export interface SkillFileNode {
+  name: string;
+  type: "file" | "dir";
+  path: string;
+  size?: number | null;
+  mtime?: number | null;
+  children?: SkillFileNode[] | null;
+}
+
+export interface SkillFilesResponse {
+  name: string;
+  category: "public" | "custom";
+  editable: boolean;
+  root: SkillFileNode;
+}
+
+export interface SkillFileContentResponse {
+  name: string;
+  category: "public" | "custom";
+  path: string;
+  mime: string;
+  size: number;
+  mtime: number;
+  is_text: boolean;
+  editable: boolean;
+  content?: string | null;
+}
+
+export interface SkillFileSaveResponse {
+  name: string;
+  path: string;
+  size: number;
+  mtime: number;
+}
+
+export async function getSkillFiles(skillName: string): Promise<SkillFilesResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/files`,
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to load skill files");
+  }
+  return response.json() as Promise<SkillFilesResponse>;
+}
+
+export async function getSkillFileContent(
+  skillName: string,
+  path: string,
+): Promise<SkillFileContentResponse> {
+  const url =
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/files/content` +
+    `?path=${encodeURIComponent(path)}`;
+  const response = await apiFetch(url);
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to read skill file");
+  }
+  return response.json() as Promise<SkillFileContentResponse>;
+}
+
+/** URL suitable for direct `<img>` / `<iframe>` / `<a download>` embedding. */
+export function getSkillFileDownloadURL(skillName: string, path: string): string {
+  return (
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/files/download` +
+    `?path=${encodeURIComponent(path)}`
+  );
+}
+
+export async function saveSkillFile(
+  skillName: string,
+  path: string,
+  content: string,
+): Promise<SkillFileSaveResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/files/content`,
+    {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ path, content }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to save skill file");
+  }
+  return response.json() as Promise<SkillFileSaveResponse>;
+}
+
+export async function deleteSkillFile(
+  skillName: string,
+  path: string,
+): Promise<void> {
+  const url =
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/files/content` +
+    `?path=${encodeURIComponent(path)}`;
+  const response = await apiFetch(url, { method: "DELETE" });
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to delete skill file");
+  }
+}
+
+export interface SkillNameZhResponse {
+  name: string;
+  name_zh: string | null;
+}
+
+export interface GenerateMissingSkillNameZhResponse {
+  generated: SkillNameZhResponse[];
+  failed: string[];
+  skipped: number;
+  /** Representative model error when some skills failed due to the LLM. */
+  model_error?: string | null;
+}
+
+/**
+ * Set a skill's Chinese display name. An empty string clears it (falls back to
+ * the English name).
+ */
+export async function setSkillNameZh(
+  skillName: string,
+  nameZh: string,
+): Promise<SkillNameZhResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/name-zh`,
+    {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ name_zh: nameZh }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to set skill Chinese name");
+  }
+  return response.json() as Promise<SkillNameZhResponse>;
+}
+
+/** Ask the LLM to generate and persist a Chinese name for a single skill. */
+export async function generateSkillNameZh(
+  skillName: string,
+  modelName?: string | null,
+): Promise<SkillNameZhResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/name-zh/generate`,
+    {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ model_name: modelName ?? null }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to generate skill Chinese name");
+  }
+  return response.json() as Promise<SkillNameZhResponse>;
+}
+
+/** One-click (admin): generate Chinese names for every skill that lacks one. */
+export async function generateMissingSkillNameZh(
+  modelName?: string | null,
+): Promise<GenerateMissingSkillNameZhResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/name-zh/generate-missing`,
+    {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ model_name: modelName ?? null }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to bulk-generate skill Chinese names");
+  }
+  return response.json() as Promise<GenerateMissingSkillNameZhResponse>;
+}
+
+export interface SetSkillDetailResponse {
+  name: string;
+  detail: string | null;
+}
+
+/**
+ * Update a skill's long-form detail text. Owner/admin for custom skills,
+ * admin-only for built-in skills (backend enforces). An empty string clears it.
+ */
+export async function setSkillDetail(
+  skillName: string,
+  detail: string,
+): Promise<SetSkillDetailResponse> {
+  const response = await apiFetch(
+    `${getBackendBaseURL()}/api/skills/${encodeURIComponent(skillName)}/detail`,
+    {
+      method: "PUT",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ detail }),
+    },
+  );
+  if (!response.ok) {
+    throw await parseDetailedError(response, "Failed to update skill detail");
+  }
+  return response.json() as Promise<SetSkillDetailResponse>;
+}

frontend-web/src/core/skills/hooks.ts

@@ -0,0 +1,379 @@
+import {
+  keepPreviousData,
+  useMutation,
+  useQuery,
+  useQueryClient,
+} from "@tanstack/react-query";
+
+import {
+  deleteBuiltinSkill,
+  deleteCustomSkill,
+  deleteSkillFile,
+  duplicateSkill,
+  enableSkill,
+  generateMissingSkillNameZh,
+  generateSkillNameZh,
+  getSkill,
+  getSkillContent,
+  getSkillFileContent,
+  getSkillFiles,
+  installSkillUpload,
+  reconcileCustomSkills,
+  saveSkillFile,
+  setSkillAlwaysOn,
+  setSkillDetail,
+  setSkillNameZh,
+  setSkillDisplay,
+  setSkillPinned,
+  setSkillFeatured,
+  setSkillPublished,
+  setSkillState,
+  validateSkillUpload,
+} from "./api";
+import type { ListSkillsParams } from "./api";
+import type { Skill, SkillState } from "./type";
+
+import { loadAllSkillsAdmin, loadSkills } from ".";
+
+// Stable empty fallback. Returning `data ?? []` inline would hand out a fresh
+// array reference on every render while the query is loading/errored, which
+// makes downstream `useMemo([skills])` / `useEffect([skills])` consumers churn
+// forever (observed as a "Maximum update depth exceeded" loop on NewAgentPage).
+const EMPTY_SKILLS: Skill[] = [];
+
+export function useSkills(params: ListSkillsParams = {}) {
+  const search = params.search ?? "";
+  const tagIds = params.tagIds ?? [];
+  const sort = params.sort ?? "time_desc";
+  const scope = params.scope;
+  const { data, isLoading, error } = useQuery({
+    queryKey: ["skills", search, [...tagIds].sort(), sort, scope ?? "all"],
+    queryFn: () => loadSkills({ search, tagIds, sort, scope }),
+    placeholderData: keepPreviousData,
+  });
+  return { skills: data ?? EMPTY_SKILLS, isLoading, error };
+}
+
+/**
+ * Admin-only listing — every skill in the system (built-in + every user's
+ * custom skills, including unpublished ones owned by other users). Backend
+ * 403s if the caller is not ``system_role === "admin"``.
+ */
+export function useAllSkillsAdmin(enabled: boolean) {
+  const { data, isLoading, error } = useQuery({
+    queryKey: ["skills", "admin", "all"],
+    queryFn: () => loadAllSkillsAdmin(),
+    enabled,
+    staleTime: 30_000,
+  });
+  return { skills: data ?? EMPTY_SKILLS, isLoading, error };
+}
+
+export function useSkillContent(skillName: string | null) {
+  return useQuery({
+    queryKey: ["skill-content", skillName],
+    queryFn: () => getSkillContent(skillName!),
+    enabled: !!skillName,
+  });
+}
+
+export function useSkill(skillName: string | null) {
+  return useQuery({
+    queryKey: ["skill", skillName],
+    queryFn: () => getSkill(skillName!),
+    enabled: !!skillName,
+  });
+}
+
+export function useSetSkillDetail() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({ skillName, detail }: { skillName: string; detail: string }) =>
+      setSkillDetail(skillName, detail),
+    onSuccess: (_data, { skillName }) => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+      void queryClient.invalidateQueries({ queryKey: ["skill", skillName] });
+    },
+  });
+}
+
+export function useSetSkillNameZh() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({ skillName, nameZh }: { skillName: string; nameZh: string }) =>
+      setSkillNameZh(skillName, nameZh),
+    onSuccess: (_data, { skillName }) => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+      void queryClient.invalidateQueries({ queryKey: ["skill", skillName] });
+    },
+  });
+}
+
+export function useGenerateSkillNameZh() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({ skillName, modelName }: { skillName: string; modelName?: string | null }) =>
+      generateSkillNameZh(skillName, modelName),
+    onSuccess: (_data, { skillName }) => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+      void queryClient.invalidateQueries({ queryKey: ["skill", skillName] });
+    },
+  });
+}
+
+export function useGenerateMissingSkillNameZh() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: (modelName?: string | null) => generateMissingSkillNameZh(modelName),
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+    },
+  });
+}
+
+export function useSetSkillDisplay() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({
+      skillName,
+      display,
+      sample,
+    }: {
+      skillName: string;
+      display: Parameters<typeof setSkillDisplay>[1];
+      sample?: Parameters<typeof setSkillDisplay>[2];
+    }) => setSkillDisplay(skillName, display, sample),
+    onSuccess: (data, { skillName }) => {
+      queryClient.setQueryData(["skill", skillName], data);
+      queryClient.setQueriesData<Skill[]>({ queryKey: ["skills"] }, (current) =>
+        Array.isArray(current)
+          ? current.map((item) => (item.name === skillName ? { ...item, ...data } : item))
+          : current,
+      );
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+      void queryClient.invalidateQueries({ queryKey: ["skill", skillName] });
+    },
+  });
+}
+
+export function useEnableSkill() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: async ({
+      skillName,
+      enabled,
+    }: {
+      skillName: string;
+      enabled: boolean;
+    }) => {
+      await enableSkill(skillName, enabled);
+    },
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+    },
+  });
+}
+
+export function useSetSkillAlwaysOn() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({ skillName, alwaysOn }: { skillName: string; alwaysOn: boolean }) =>
+      setSkillAlwaysOn(skillName, alwaysOn),
+    onSuccess: (data, { skillName }) => {
+      queryClient.setQueryData(["skill", skillName], data);
+      queryClient.setQueriesData<Skill[]>({ queryKey: ["skills"] }, (current) =>
+        Array.isArray(current)
+          ? current.map((item) => (item.name === skillName ? { ...item, ...data } : item))
+          : current,
+      );
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+      void queryClient.invalidateQueries({ queryKey: ["skill", skillName] });
+    },
+  });
+}
+
+export function useValidateSkillUpload() {
+  return useMutation({
+    mutationFn: ({
+      file,
+      targetCategory,
+    }: {
+      file: File;
+      targetCategory?: "public" | "custom";
+    }) => validateSkillUpload(file, targetCategory),
+  });
+}
+
+export function useInstallSkillUpload() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({ file, overwrite, targetCategory }: { file: File; overwrite?: boolean; targetCategory?: "public" | "custom" }) =>
+      installSkillUpload(file, { overwrite, targetCategory }),
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+    },
+  });
+}
+
+export function useReconcileCustomSkills() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: () => reconcileCustomSkills(),
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+    },
+  });
+}
+
+export function useSetSkillPublished() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({
+      skillName,
+      published,
+      squareId,
+    }: {
+      skillName: string;
+      published: boolean;
+      squareId?: string;
+    }) => setSkillPublished(skillName, published, squareId),
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+      // Unpublishing triggers the backend cascade that strips this skill
+      // from every referencing agent's config.yaml, so the cached agent
+      // list (agent.skills) is now stale.
+      void queryClient.invalidateQueries({ queryKey: ["agents"] });
+    },
+  });
+}
+
+export function useSetSkillState() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({ skillName, state }: { skillName: string; state: SkillState }) =>
+      setSkillState(skillName, state),
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+    },
+  });
+}
+
+export function useSetSkillPinned() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({ skillName, pinned }: { skillName: string; pinned: boolean }) =>
+      setSkillPinned(skillName, pinned),
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+    },
+  });
+}
+
+export function useSetSkillFeatured() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({ skillName, featured }: { skillName: string; featured: boolean }) =>
+      setSkillFeatured(skillName, featured),
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+    },
+  });
+}
+
+export function useDuplicateSkill() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({
+      sourceName,
+      newName,
+    }: {
+      sourceName: string;
+      newName: string;
+    }) => duplicateSkill(sourceName, newName),
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+    },
+  });
+}
+
+export function useDeleteCustomSkill() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: (skillName: string) => deleteCustomSkill(skillName),
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+      // Deletion triggers the backend cascade that strips this skill from
+      // every referencing agent's config.yaml, so the cached agent list
+      // (agent.skills) is now stale.
+      void queryClient.invalidateQueries({ queryKey: ["agents"] });
+    },
+  });
+}
+
+export function useSkillFiles(skillName: string | null) {
+  return useQuery({
+    queryKey: ["skill-files", skillName],
+    queryFn: () => getSkillFiles(skillName!),
+    enabled: !!skillName,
+  });
+}
+
+export function useSkillFileContent(skillName: string | null, path: string | null) {
+  return useQuery({
+    queryKey: ["skill-file", skillName, path],
+    queryFn: () => getSkillFileContent(skillName!, path!),
+    enabled: !!skillName && !!path,
+  });
+}
+
+export function useSaveSkillFile() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({
+      skillName,
+      path,
+      content,
+    }: {
+      skillName: string;
+      path: string;
+      content: string;
+    }) => saveSkillFile(skillName, path, content),
+    onSuccess: (_data, { skillName, path }) => {
+      void queryClient.invalidateQueries({ queryKey: ["skill-files", skillName] });
+      void queryClient.invalidateQueries({ queryKey: ["skill-file", skillName, path] });
+      // SKILL.md edits also affect the source-tab view and skill metadata.
+      if (path === "SKILL.md") {
+        void queryClient.invalidateQueries({ queryKey: ["skill-content", skillName] });
+        void queryClient.invalidateQueries({ queryKey: ["skill", skillName] });
+      }
+    },
+  });
+}
+
+export function useDeleteSkillFile() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({
+      skillName,
+      path,
+    }: {
+      skillName: string;
+      path: string;
+    }) => deleteSkillFile(skillName, path),
+    onSuccess: (_data, { skillName, path }) => {
+      void queryClient.invalidateQueries({ queryKey: ["skill-files", skillName] });
+      void queryClient.removeQueries({ queryKey: ["skill-file", skillName, path] });
+    },
+  });
+}
+
+export function useDeleteBuiltinSkill() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: deleteBuiltinSkill,
+    onSuccess: () => {
+      void queryClient.invalidateQueries({ queryKey: ["skills"] });
+      void queryClient.invalidateQueries({ queryKey: ["admin-skills"] });
+    },
+  });
+}

frontend-web/src/core/skills/index.ts

@@ -0,0 +1,3 @@
+export * from "./api";
+export * from "./type";
+export * from "./utils";

frontend-web/src/core/skills/type.ts

@@ -0,0 +1,88 @@
+import type { Tag } from "@/core/tags";
+
+export type SkillState = "active" | "stale" | "archived";
+export type SkillSource = "agent" | "upload";
+export type SkillDisplayMode = "citation" | "list" | "table" | "card" | "custom" | "none";
+export type SkillLinkStrategy = "auto" | "url" | "id";
+export type SkillCitationToken = "doubleBracket" | "singleBracket";
+
+export interface SkillDisplayTableColumn {
+  label: string;
+  field: string;
+  width?: string | null;
+}
+
+export interface SkillDisplayResultSetConfig {
+  key: string;
+  label?: string | null;
+  enabled: boolean;
+  mode: SkillDisplayMode;
+  resultPath: string;
+  titleField?: string | null;
+  summaryField?: string | null;
+  contentField?: string | null;
+  sourceField?: string | null;
+  authorField?: string | null;
+  typeField?: string | null;
+  timeField?: string | null;
+  scoreField?: string | null;
+  urlField?: string | null;
+  idField?: string | null;
+  linkStrategy: SkillLinkStrategy;
+  linkTemplate?: string | null;
+  citationToken: SkillCitationToken;
+  maxItems: number;
+  customInstruction: string;
+  tableColumns?: SkillDisplayTableColumn[];
+}
+
+export interface SkillDisplayConfig extends Omit<SkillDisplayResultSetConfig, "key" | "label"> {
+  resultSets?: SkillDisplayResultSetConfig[];
+}
+
+export interface Skill {
+  name: string;
+  // 用户配置或大模型生成的中文名称;为空时界面回退到使用英文 name。
+  name_zh?: string | null;
+  description: string;
+  category: string;
+  license: string;
+  enabled: boolean;
+  // 常驻(免压缩) —— 开启技能压缩后,标记此项的技能仍以全文形式注入系统提示词,
+  // 而非被压缩成索引 + search_skills 检索。
+  always_on?: boolean;
+  // Multitenancy fields — owner_user_id is null for built-in / public skills.
+  // ``builtin`` mirrors that nullness for explicit branching on the UI side.
+  owner_user_id?: string | null;
+  // Publisher username (email local part) resolved from owner_user_id by the
+  // backend. Null for built-in / public skills with no human owner.
+  owner_username?: string | null;
+  published?: boolean;
+  /** Publish square (发布广场) the skill is filed under once published. "" = default square. */
+  square_id?: string;
+  builtin?: boolean;
+  // True when this skill is in the caller's 我的技能 collection — e.g. granted by
+  // their 岗位 (position). Owned skills are detected separately via owner_user_id.
+  favorited?: boolean;
+  // Long-form detail/introduction authored by the skill owner or an admin.
+  // Distinct from ``description`` (parsed from SKILL.md); supports Markdown.
+  detail?: string | null;
+  display?: SkillDisplayConfig | null;
+  displaySample?: Record<string, unknown> | null;
+  // Lifecycle fields — populated by curator. The main list filters on these
+  // so stale / archived skills appear in dedicated drawers instead.
+  state?: SkillState;
+  pinned?: boolean;
+  // 置顶 —— 仅影响列表排序,与 pinned(防自动归档)正交。
+  featured?: boolean;
+  // Provenance — distinguishes skills created by an agent during a chat
+  // (``agent``) from skills the user uploaded / authored manually (``upload``).
+  source?: SkillSource;
+  created_at?: string | null;
+  last_activity_at?: string | null;
+  use_count?: number;
+  view_count?: number;
+  patch_count?: number;
+  // Tags assigned to this skill (admin-managed).
+  tags?: Tag[];
+}

frontend-web/src/core/skills/utils.ts

@@ -0,0 +1,22 @@
+import type { Skill } from "./type";
+
+/**
+ * 智能体技能白名单的候选池过滤。
+ *
+ * 编辑智能体时，普通用户只能从「内置技能 + 自己的技能 + 已发布到广场的
+ * 技能」里挑选，其他用户未发布的自定义技能不可选；管理员仍可使用全部
+ * 已启用技能。
+ */
+export function isAgentSkillCandidate(
+  skill: Skill,
+  currentUserId: string | null,
+  isAdmin = false,
+): boolean {
+  if (!skill.enabled) return false;
+  if (isAdmin) return true;
+  if ((skill.state ?? "active") !== "active") return false;
+  const builtin = skill.builtin ?? !skill.owner_user_id;
+  const ownedByMe =
+    !!skill.owner_user_id && skill.owner_user_id === currentUserId;
+  return builtin || ownedByMe || !!skill.published;
+}

frontend-web/src/eightbf-roundtable/components/EightBfStep1Workspace.tsx

@@ -31,6 +31,7 @@ import {
   syncPositionOptionsFromRoles,
 } from "@/position-roundtable/lib/positions";
 import type { DraftStep3SummarySnapshot } from "@/roundtable-planning/api/drafts";
+import { EIGHTBF_INTENT_AGENT_ID } from "@/roundtable-planning/api/intent";
 import type { ThreadIdMap } from "@/roundtable-planning/api/multi-agent";
 import "@/position-roundtable/styles/intent-understanding.css";
 import type { Step1PanelProps } from "@/roundtable-planning/components/Step1Panel";
@@ -96,6 +97,8 @@ export function EightBfStep1Panel({
 }: EightBfStep1PanelProps) {
   const [rightPanelOpen, setRightPanelOpen] = useState(true);
   const [rightTab, setRightTab] = useState<"agents" | "results">("agents");
+  // 第一步内置任务理解智能体：其「常用问题」显示在输入框左上角，点击直接发送。
+  const { agent: intentAgent } = useAgent(EIGHTBF_INTENT_AGENT_ID);
 
   const openPanel = (tab: "agents" | "results") => {
     if (rightPanelOpen && rightTab === tab) {
@@ -167,6 +170,8 @@ export function EightBfStep1Panel({
             modelOptions={modelOptions}
             selectedModel={selectedModel}
             onSelectedModelChange={onSelectedModelChange}
+            commonQuestions={intentAgent?.common_questions ?? null}
+            sendCommonQuestionOnClick
           />
         )}
         intentReady={intentReady}
@@ -331,8 +336,26 @@ function EightBfAgentWorkspace({
     summaryReport: step3Summary,
   });
 
+  // 第一步绑定的内置任务理解智能体：不在业务链席位里，固定置顶展示，
+  // 供管理员在本面板查看 / 配置其提示词与常用问题。
+  const intentAgent = useMemo(
+    () => agents.find((agent) => agent.id === EIGHTBF_INTENT_AGENT_ID) ?? null,
+    [agents],
+  );
+  const intentAgentMatchesKeyword = useMemo(() => {
+    const keyword = agentKeyword.trim().toLowerCase();
+    if (!keyword) return true;
+    const name = (intentAgent?.name || "8BF 任务理解").toLowerCase();
+    const description = (intentAgent?.description || "").toLowerCase();
+    return name.includes(keyword) || description.includes(keyword);
+  }, [agentKeyword, intentAgent]);
+
   useEffect(() => {
-    if (selectedAgentId && !seats.some((seat) => seat.agent_id === selectedAgentId)) {
+    if (
+      selectedAgentId &&
+      selectedAgentId !== EIGHTBF_INTENT_AGENT_ID &&
+      !seats.some((seat) => seat.agent_id === selectedAgentId)
+    ) {
       setSelectedAgentId(null);
     }
   }, [seats, selectedAgentId]);
@@ -412,7 +435,12 @@ function EightBfAgentWorkspace({
           ) : (
             <AgentDetailPanel
               agent={selectedAgent}
-              agentName={selectedAgent?.name ?? selectedSeat?.name ?? ""}
+              agentName={
+                selectedAgent?.name
+                ?? (selectedAgentId === EIGHTBF_INTENT_AGENT_ID
+                  ? "8BF 任务理解"
+                  : selectedSeat?.name ?? "")
+              }
               onAskQuestion={() => toast.info("请在中间对话区向任务理解智能体发起问题")}
               showRecentChats={false}
               avatarSrc={agentConfigLogo}
@@ -468,7 +496,7 @@ function EightBfAgentWorkspace({
         </header>
         {isLoading && seats.length > 0 ? (
           <div className="eightbf-roundtable__agent-selector-empty">正在加载智能体…</div>
-        ) : filteredSeatAgents.length === 0 ? (
+        ) : filteredSeatAgents.length === 0 && !intentAgentMatchesKeyword ? (
           <div className="eightbf-roundtable__agent-selector-empty">
             <p>{seatAgents.length ? "没有符合条件的智能体" : "当前岗位尚未配置智能体"}</p>
             {!seatAgents.length && canConfigureChain && (
@@ -477,6 +505,13 @@ function EightBfAgentWorkspace({
           </div>
         ) : (
           <div className="eightbf-roundtable__agent-selector-list">
+            {intentAgentMatchesKeyword && (
+              <EightBfAgentCard
+                name={intentAgent?.name || "8BF 任务理解"}
+                description={intentAgent?.description || "第一步「意图识别」的内置任务理解智能体"}
+                onSelect={() => setSelectedAgentId(EIGHTBF_INTENT_AGENT_ID)}
+              />
+            )}
             {filteredSeatAgents.map(({ seat, agent }) => (
               <EightBfAgentRow
                 key={seat.agent_id}
@@ -503,6 +538,19 @@ function EightBfAgentRow({
 }) {
   const name = agent?.name || seat.name || "未命名智能体";
   const description = agent?.description || seat.description || "智能体助手";
+  return <EightBfAgentCard name={name} description={description} onSelect={onSelect} />;
+}
+
+/** 智能体列表卡片：席位智能体与置顶的内置任务理解智能体共用同一展示结构。 */
+function EightBfAgentCard({
+  name,
+  description,
+  onSelect,
+}: {
+  name: string;
+  description: string;
+  onSelect: () => void;
+}) {
   const avatarText = firstAgentAvatarChar(name);
   return (
     <button type="button" className="eightbf-roundtable__agent-selector-card" onClick={onSelect}>

frontend-web/src/eightbf-roundtable/pages/EightBfRoundtablePage.tsx

@@ -3,6 +3,7 @@ import { useSearchParams } from "react-router-dom";
 import { Loader2, Save, X } from "lucide-react";
 
 import { getAccountDisplayName, getStoredAuth } from "@/core/auth";
+import { EIGHTBF_INTENT_AGENT_ID } from "@/roundtable-planning/api/intent";
 import { EightBfTaskCard } from "@/eightbf-roundtable/components/EightBfTaskCard";
 import {
   EightBfStep1Panel,
@@ -76,6 +77,9 @@ export default function EightBfRoundtablePage({ compactChrome = false }: { compa
   const [selectedChainId, setSelectedChainId] = useState("");
   const [isSaving, setIsSaving] = useState(false);
   const [error, setError] = useState<string | null>(null);
+  // 第二步「人工干预」：开启后每走一步暂停等待用户输入；关闭自动推进。会话内状态，
+  // 新建会话 / 刷新后回到默认关闭（与现状一致）。
+  const [step2Intervention, setStep2Intervention] = useState(false);
   const configuredChain = useMemo(
     () => chains.find((chain) => chain.id === mapping?.chainId) ?? null,
     [chains, mapping?.chainId],
@@ -260,6 +264,12 @@ export default function EightBfRoundtablePage({ compactChrome = false }: { compa
       // 无 taskId 进入时每次都开全新会话（不回显 anonymous 键下的最新草稿）；带 taskId 保持回显现状。
       skipInitialDraftRestore={!taskId}
       taskContext={taskContext as Record<string, unknown> | null}
+      // 第一步绑定 8BF 专属内置意图智能体（roundtable-intent 的同源复制，
+      // 提示词独立可调优），不再与通用多智能体会商共用一个 SOUL。
+      intentAgentId={EIGHTBF_INTENT_AGENT_ID}
+      // 页头「人工干预」开关：控制第二步是否每步暂停等待用户输入。
+      step2InterventionEnabled={step2Intervention}
+      onStep2InterventionChange={setStep2Intervention}
       entryRailTaskId={taskId || null}
       entryRailTaskCardSlot={taskId ? (
         <EightBfTaskCard

frontend-web/src/open-canvas/pages/AIWritingPage.tsx

@@ -142,7 +142,7 @@ const DRAFT_ID = 'aw-draft';
 const AI_WRITING_AGENT_MODULE_KEY = 'ai-writing';
 const AI_WRITING_AGENT_MODULE_TITLE = 'AI 写作智能体';
 const AI_WRITING_LAST_AGENT_KEY = `agent-module:${AI_WRITING_AGENT_MODULE_KEY}:last-agent`;
-const SHOW_NOTEBOOK_SPACE_ENTRIES = false;
+const SHOW_NOTEBOOK_SPACE_ENTRIES = true;
 
 // Match the original canvas page proportions (chat=35, artifact=65)
 const DEFAULT_LAYOUT = { [WRITING_ID]: 35, [DRAFT_ID]: 65 };

frontend-web/src/pages/AgentsOrbitPage.tsx

@@ -0,0 +1,247 @@
+import { useEffect, useMemo, useRef, useState } from "react";
+import { Bot } from "lucide-react";
+import { useRouter } from "next/navigation";
+
+import { useAgents } from "@/core/agents";
+import type { Agent } from "@/core/agents";
+import { themeToBaseMode, useTheme } from "@/contexts/ThemeContext";
+
+import "@/styles/agents-orbit.css";
+import "@/styles/agents-orbit-dark.css";
+import "@/styles/agents-orbit-light.css";
+
+/** 转盘旋转过渡时长（ms），与 CSS 中 .ao-type-arm 的 transition 保持一致。 */
+const DIAL_TRANSITION_MS = 850;
+
+/** 一个类型 = 一个标签（取智能体的第一个标签；无标签归入「未分类」）。 */
+interface AgentGroup {
+  id: string;
+  name: string;
+  agents: Agent[];
+}
+
+function groupAgentsByType(agents: Agent[]): AgentGroup[] {
+  const map = new Map<string, AgentGroup>();
+  for (const agent of agents) {
+    const tag = agent.tags?.[0];
+    const id = tag?.id ?? "__uncategorized__";
+    let group = map.get(id);
+    if (!group) {
+      group = { id, name: tag?.name ?? "未分类", agents: [] };
+      map.set(id, group);
+    }
+    group.agents.push(agent);
+  }
+  return Array.from(map.values()).sort((a, b) =>
+    a.name.localeCompare(b.name, "zh-Hans-CN"),
+  );
+}
+
+const mod = (value: number, n: number): number => ((value % n) + n) % n;
+
+/**
+ * 密码转盘模型：
+ * - n 个类型始终均匀占据右半圆的 n 个固定槽位（间距恒定，不随点击改变）；
+ * - 激活槽位固定为最贴近正右侧中间的槽，被点击的类型旋转到该槽；
+ * - 点击上半部分 → 转盘顺时针转（k > 0），下半部分 → 逆时针（k < 0）；
+ * - 超出半圆端点的标签沿“背面长路”补位（角度 ±360 扫过左半圆，途中半透明），
+ *   动画结束后角度归一化回槽位角（同一位置 mod 360，无视觉跳变）。
+ */
+export default function AgentsOrbitPage() {
+  const router = useRouter();
+  const { theme } = useTheme();
+  const isDark = themeToBaseMode(theme) === "dark";
+  const { agents, isLoading } = useAgents();
+  const groups = useMemo(() => groupAgentsByType(agents), [agents]);
+
+  const [activeId, setActiveId] = useState<string | null>(null);
+  /** 转盘动画期间的非归一化角度（含 ±360 长路，长路全程在左半圆不可视区内）。 */
+  const [anim, setAnim] = useState<Record<string, number> | null>(null);
+  /** 归一化帧：瞬移到等价角度时禁用过渡。 */
+  const [frozen, setFrozen] = useState(false);
+  const timerRef = useRef<number | null>(null);
+
+  useEffect(() => {
+    return () => {
+      if (timerRef.current != null) window.clearTimeout(timerRef.current);
+    };
+  }, []);
+
+  if (isLoading) {
+    return (
+      <div className={`agents-orbit-page ${isDark ? "ao-dark" : "ao-light"}`}>
+        <div className="ao-state">
+          <div className="ao-spinner" />
+          <span>正在加载智能体…</span>
+        </div>
+      </div>
+    );
+  }
+
+  if (groups.length === 0) {
+    return (
+      <div className={`agents-orbit-page ${isDark ? "ao-dark" : "ao-light"}`}>
+        <div className="ao-state">
+          <span>还没有可展示的智能体</span>
+        </div>
+      </div>
+    );
+  }
+
+  const n = groups.length;
+  const midSlot = Math.floor((n - 1) / 2);
+  const step = 180 / n;
+  /** 槽位 j 的世界角：-90°(正上) → +90°(正下)，均布在右半圆开区间内。 */
+  const slotAngle = (slot: number) => -90 + (slot + 0.5) * step;
+
+  const requestedIdx = activeId
+    ? groups.findIndex((g) => g.id === activeId)
+    : -1;
+  const activeIdx = requestedIdx >= 0 ? requestedIdx : midSlot;
+  const active = groups[activeIdx];
+  /** 静态分组数组的旋转量：group(gi) 落在槽位 (gi + shift) mod n，激活组恰在 midSlot。 */
+  const shift = midSlot - activeIdx;
+  const groupAtSlot = (slot: number) => groups[mod(slot - shift, n)];
+
+  const handleTypeClick = (slot: number) => {
+    const group = groupAtSlot(slot);
+    if (group.id === active.id) return;
+
+    const k = midSlot - slot; // 正=顺时针向下转，负=逆时针向上转
+    const angles: Record<string, number> = {};
+    for (let j = 0; j < n; j++) {
+      const g = groupAtSlot(j);
+      const nextSlot = j + k;
+      let angle = slotAngle(mod(nextSlot, n));
+      if (nextSlot >= n) {
+        angle += 360; // 从底部滑出（进入左半圆不可视区），绕背回顶部补位
+      } else if (nextSlot < 0) {
+        angle -= 360; // 从顶部滑出（进入左半圆不可视区），绕背回底部补位
+      }
+      angles[g.id] = angle;
+    }
+
+    setActiveId(group.id);
+    setAnim(angles);
+    if (timerRef.current != null) window.clearTimeout(timerRef.current);
+    timerRef.current = window.setTimeout(() => {
+      // 动画结束：归一化长路角度（同一位置 mod 360），下一帧恢复过渡。
+      setFrozen(true);
+      setAnim(null);
+      requestAnimationFrame(() =>
+        requestAnimationFrame(() => setFrozen(false)),
+      );
+    }, DIAL_TRANSITION_MS + 60);
+  };
+
+  const handleAgentClick = (agent: Agent) => {
+    router.push(`/page/workspace/agents/${agent.id}/chats/new`);
+  };
+
+  const themeClass = isDark ? "ao-dark" : "ao-light";
+
+  return (
+    <div
+      className={`agents-orbit-page ${themeClass}${n >= 8 ? " ao-crowded" : ""}`}
+    >
+      <div className="ao-header">
+        <span className="ao-header-title">智能体 · 星轨转盘</span>
+        <button
+          type="button"
+          className="ao-header-link"
+          onClick={() => router.push("/page/workspace/agents")}
+        >
+          经典视图
+        </button>
+      </div>
+
+      {/* 左侧星轨舞台：中心圆盘 + 中心图片 + 三层圆环 + 卫星 + 类型转盘 */}
+      <div className="ao-stage">
+        <div className="ao-root">
+          <div className="ao-ring ao-ring--ticks" />
+          <div className="ao-ring ao-ring--dash" />
+          <div className="ao-ring ao-ring--arc" />
+          <div className="ao-sat-orbit">
+            <div className="ao-sat" />
+          </div>
+
+          <div className="ao-core">
+            <div className="ao-core-img">
+              <Bot size={60} strokeWidth={1.4} />
+            </div>
+          </div>
+
+          {/* 转盘容器：左半圆被裁剪为不可视区域，绕背旋转不可见 */}
+          <div className="ao-types">
+            {groups.map((group, gi) => {
+              const slot = mod(gi + shift, n);
+              const angle = anim
+                ? (anim[group.id] ?? slotAngle(slot))
+                : slotAngle(slot);
+              const isActive = group.id === active.id;
+              const transitionStyle = frozen ? { transition: "none" } : {};
+              return (
+                <div
+                  key={group.id}
+                  className="ao-type-arm"
+                  style={{
+                    transform: `rotate(${angle}deg)`,
+                    ...transitionStyle,
+                  }}
+                >
+                  <button
+                    type="button"
+                    className={`ao-type-anchor${isActive ? " is-active" : ""}`}
+                    style={{
+                      transform: `translate(-50%, -50%) rotate(${-angle}deg)`,
+                      ...transitionStyle,
+                    }}
+                    onClick={() => handleTypeClick(slot)}
+                    title={`${group.name} · ${group.agents.length} 个智能体`}
+                  >
+                    <span
+                      className="ao-type-chip"
+                      style={{ animationDelay: `${-gi * 0.8}s` }}
+                    >
+                      {group.name}
+                    </span>
+                  </button>
+                </div>
+              );
+            })}
+          </div>
+        </div>
+      </div>
+
+      {/* 右侧面板：当前类型下的智能体（圆角正方形图标 + 名称），与舞台 1:1 */}
+      <aside className="ao-panel">
+        <div className="ao-panel-head">
+          <div className="ao-panel-title">{active.name}</div>
+          <div className="ao-panel-sub">共 {active.agents.length} 个智能体</div>
+        </div>
+        {active.agents.length > 0 ? (
+          <div className="ao-agents" key={active.id}>
+            {active.agents.map((agent) => (
+              <button
+                key={agent.id}
+                type="button"
+                className="ao-agent"
+                onClick={() => handleAgentClick(agent)}
+                title={agent.description || agent.name}
+              >
+                <span className="ao-agent-icon">
+                  {(agent.name || "?").trim().charAt(0).toUpperCase()}
+                </span>
+                <span className="ao-agent-name">{agent.name}</span>
+              </button>
+            ))}
+          </div>
+        ) : (
+          <div className="ao-state">
+            <span>该类型下暂时没有智能体</span>
+          </div>
+        )}
+      </aside>
+    </div>
+  );
+}

frontend-web/src/pages/OfficialDeerFlowEntry.tsx

@@ -5,8 +5,10 @@ import {
   ChevronDown,
   CircleHelp,
   Clock3,
+  GitBranch,
   LoaderCircle,
   MessageSquareText,
+  Network,
   PanelLeftClose,
   Pause,
   Play,
@@ -41,10 +43,25 @@ import {
   OFFICIAL_DEERFLOW_HOME_ROUTE,
   OFFICIAL_DEERFLOW_ROUTE_PREFIX,
 } from "@/core/official-deerflow/routes";
+import { Toaster } from "@/components/ui/sonner";
+
+import MultiAgentTasksPage from "./official-multi-agent-tasks/MultiAgentTasksPage";
+import TaskEditPage from "./official-multi-agent-tasks/TaskEditPage";
+import TaskAnalysisPage from "./official-task-analysis/TaskAnalysisPage";
+import TaskAnalysisDetailPage from "./official-task-analysis/TaskAnalysisDetailPage";
 
 import "./official-deerflow-entry.css";
 
-type OfficialPage = "chat" | "agents" | "agent-chat" | "scheduled" | "settings";
+type OfficialPage =
+  | "chat"
+  | "agents"
+  | "agent-chat"
+  | "scheduled"
+  | "multi-agent-tasks"
+  | "multi-agent-edit"
+  | "task-analysis"
+  | "task-analysis-detail"
+  | "settings";
 
 function pageFromPath(pathname: string): OfficialPage {
   const path = pathname.slice(OFFICIAL_DEERFLOW_ROUTE_PREFIX.length);
@@ -53,6 +70,13 @@ function pageFromPath(pathname: string): OfficialPage {
   }
   if (path.startsWith("/workspace/agents")) return "agents";
   if (path.startsWith("/workspace/scheduled-tasks")) return "scheduled";
+  if (path.startsWith("/workspace/multi-agent-tasks/new")) return "multi-agent-edit";
+  if (/^\/workspace\/multi-agent-tasks\/[^/]+\/edit$/.test(path)) {
+    return "multi-agent-edit";
+  }
+  if (path.startsWith("/workspace/multi-agent-tasks")) return "multi-agent-tasks";
+  if (/^\/workspace\/task-analysis\/[^/]+/.test(path)) return "task-analysis-detail";
+  if (path.startsWith("/workspace/task-analysis")) return "task-analysis";
   if (path.startsWith("/workspace/settings")) return "settings";
   return "chat";
 }
@@ -78,6 +102,33 @@ function parseScheduledTaskId(pathname: string) {
   return match ? decodeURIComponent(match[1]!) : null;
 }
 
+function parseMultiAgentTaskId(pathname: string) {
+  const match = pathname.match(/\/workspace\/multi-agent-tasks\/([^/]+)/);
+  const raw = match ? decodeURIComponent(match[1]!) : null;
+  // 编辑页路由（new / :id/edit）不是任务详情，不驱动列表页的抽屉。
+  if (!raw || raw === "new") return null;
+  return raw;
+}
+
+function parseTaskAnalysisId(pathname: string) {
+  const match = pathname.match(/\/workspace\/task-analysis\/([^/]+)/);
+  return match ? decodeURIComponent(match[1]!) : null;
+}
+
+function parseMultiAgentEditTarget(pathname: string): {
+  mode: "create" | "edit";
+  taskId: string | null;
+} {
+  const path = pathname.slice(OFFICIAL_DEERFLOW_ROUTE_PREFIX.length);
+  if (path.startsWith("/workspace/multi-agent-tasks/new")) {
+    return { mode: "create", taskId: null };
+  }
+  const match = path.match(/^\/workspace\/multi-agent-tasks\/([^/]+)\/edit$/);
+  return match
+    ? { mode: "edit", taskId: decodeURIComponent(match[1]!) }
+    : { mode: "create", taskId: null };
+}
+
 function readableError(error: unknown, fallback: string) {
   return error instanceof Error && error.message ? error.message : fallback;
 }
@@ -658,6 +709,18 @@ export default function OfficialDeerFlowEntry() {
     () => parseScheduledTaskId(location.pathname),
     [location.pathname],
   );
+  const multiAgentTaskId = useMemo(
+    () => parseMultiAgentTaskId(location.pathname),
+    [location.pathname],
+  );
+  const multiAgentEditTarget = useMemo(
+    () => parseMultiAgentEditTarget(location.pathname),
+    [location.pathname],
+  );
+  const taskAnalysisId = useMemo(
+    () => parseTaskAnalysisId(location.pathname),
+    [location.pathname],
+  );
   const { data: threads = [], isLoading: threadsLoading, error: threadsError } =
     useThreads({
       limit: 20,
@@ -672,6 +735,8 @@ export default function OfficialDeerFlowEntry() {
     { icon: MessageSquareText, label: "对话", path: `${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/chats/new`, pages: ["chat"] },
     { icon: Bot, label: "智能体", path: `${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/agents`, pages: ["agents", "agent-chat"] },
     { icon: CalendarClock, label: "定时任务", path: `${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/scheduled-tasks`, pages: ["scheduled"] },
+    { icon: Network, label: "多智能体定时任务", path: `${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/multi-agent-tasks`, pages: ["multi-agent-tasks", "multi-agent-edit"] },
+    { icon: GitBranch, label: "任务分析定时", path: `${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/task-analysis`, pages: ["task-analysis", "task-analysis-detail"] },
   ] as const;
 
   return (
@@ -769,8 +834,23 @@ export default function OfficialDeerFlowEntry() {
           />
         )}
         {page === "scheduled" && <ScheduledTasksPage selectedTaskId={scheduledTaskId} />}
+        {page === "multi-agent-tasks" && (
+          <MultiAgentTasksPage routeTaskId={multiAgentTaskId} />
+        )}
+        {page === "multi-agent-edit" && (
+          <TaskEditPage
+            key={multiAgentEditTarget.taskId ?? "create"}
+            mode={multiAgentEditTarget.mode}
+            taskId={multiAgentEditTarget.taskId}
+          />
+        )}
+        {page === "task-analysis" && <TaskAnalysisPage />}
+        {page === "task-analysis-detail" && (
+          <TaskAnalysisDetailPage routeTaskId={taskAnalysisId} />
+        )}
         {page === "settings" && <SettingsPage />}
       </div>
+      <Toaster position="top-center" />
     </main>
   );
 }

frontend-web/src/pages/TaskButtonsPage.tsx

@@ -309,7 +309,24 @@ export default function TaskButtonsPage() {
                               className="h-8 w-24"
                               title="发给宿主的参数名（默认 id）"
                             />
-                            <span className="text-muted-foreground">= 任务id（数字）</span>
+                            <span className="text-muted-foreground">=</span>
+                            {/* 行动分析(xdfx)：行动id 与任务id 不同，可选发哪种；其它业务固定任务id */}
+                            {tab === "xdfx" ? (
+                              <Select
+                                value={r.idKind}
+                                onValueChange={(v) => update(r.id, { idKind: v as TaskButton["idKind"] })}
+                              >
+                                <SelectTrigger className="h-8 w-24">
+                                  <SelectValue />
+                                </SelectTrigger>
+                                <SelectContent>
+                                  <SelectItem value="task">任务id</SelectItem>
+                                  <SelectItem value="action">行动id</SelectItem>
+                                </SelectContent>
+                              </Select>
+                            ) : (
+                              <span className="text-muted-foreground">任务id（数字）</span>
+                            )}
                           </div>
                         )}
                         <span className="text-muted-foreground w-full text-[11px]">

frontend-web/src/pages/WorkspaceRoutes.tsx

@@ -39,6 +39,7 @@ import ChatPage from "./ChatPage";
 import ArtifactLibraryPage from "./ArtifactLibraryPage";
 import IframeChatPage from "./IframeChatPage";
 import AgentsPage from "./AgentsPage";
+import AgentsOrbitPage from "./AgentsOrbitPage";
 import AgentChatPage from "./AgentChatPage";
 import SentimentAgentChatPage from "./SentimentAgentChatPage";
 import NewAgentPage from "./NewAgentPage";
@@ -620,6 +621,16 @@ export default function WorkspaceRoutes({
             </WorkspaceLayout>
           }
         />
+        {/* 智能体星轨视图：/agents 的另一种展示形式（手写 CSS 圆环星轨布局）。
+            字面量路由，置于 agents/:agent_id/* 参数路由之前。 */}
+        <Route
+          path="agents/orbit"
+          element={
+            <WorkspaceLayout {...layoutProps}>
+              <AgentsOrbitPage />
+            </WorkspaceLayout>
+          }
+        />
         {/* 舆情分析虚拟智能体：静态段优先于 :agent_id 命中，走 Gateway 转发
             外部 AG-UI 网关的专用对话页（不在后端 /api/agents 落库）。 */}
         <Route

frontend-web/src/pages/deep-research/WorkflowConfigDialog.tsx

@@ -8,8 +8,10 @@
  *    全新部署自动播种「通用分析结构 / 递进式论证结构」），支持 启用/停用 筛选
  *    与 标题 筛选；卡片上可 switch 启停、编辑、删除（二次确认），并显示
  *    最后修改时间与创建用户名；新增时不填状态（服务端强制默认启用）。
- *  - 详情视图：点击卡片即「选择」该结构（内容回填到高级配置的报告结构
+ *    **点击卡片即「选择」该结构（内容回填到高级配置的报告结构 custom_outline，
- *    custom_outline），弹窗切换为展示 标题 + Markdown 大纲预览，右上角「返回」回列表。
+ *    卡片留在列表中高亮为当前使用，不进入详情）**。
+ *  - 详情视图：通过卡片右上角「查看大纲结构」眼睛图标进入，展示 标题 + Markdown
+ *    大纲预览，右上角「返回」回列表。
  */
 
 import { useEffect, useMemo, useState } from "react";
@@ -238,8 +240,9 @@ export function WorkflowConfigDialog({
       toast.error(`「${structure.title}」已停用，请先启用后再选用`);
       return;
     }
+    // 点击卡片只做「选择」：内容回填后留在列表视图（卡片高亮为当前使用），
+    // 详情视图通过卡片上的「查看大纲结构」眼睛图标进入。
     onApply?.(structure.content, structure.structureMode, structure.retrievalDirections, structure.retrievalSkills);
-    setDetailId(structure.id);
     toast.success(`已选择报告结构「${structure.title}」，并填充到报告结构配置`);
   };
 
@@ -325,7 +328,7 @@ export function WorkflowConfigDialog({
                     </h2>
                   </div>
                   <p className="mt-1 text-xs text-slate-500 dark:text-slate-400 sm:text-[13px]">
-                    点击卡片选择报告结构（自动填充到报告结构配置）；可启停、编辑、删除。
+                    点击卡片选择报告结构（自动填充到报告结构配置）；点击卡片上的眼睛图标查看大纲详情；可启停、编辑、删除。
                   </p>
                 </div>
                 <Button
@@ -501,7 +504,7 @@ export function WorkflowConfigDialog({
                           key={structure.id}
                           structure={structure}
                           applied={structure.id === appliedStructureId}
-                          onOpen={() => handleSelect(structure)}
+                          onSelect={() => handleSelect(structure)}
                           onPreview={(event) => {
                             event.stopPropagation();
                             setDetailId(structure.id);
@@ -750,7 +753,7 @@ export function WorkflowConfigDialog({
 function StructureCard({
   structure,
   applied,
-  onOpen,
+  onSelect,
   onPreview,
   onDuplicate,
   onEdit,
@@ -760,7 +763,7 @@ function StructureCard({
 }: {
   structure: ReportStructure;
   applied: boolean;
-  onOpen: () => void;
+  onSelect: () => void;
   onPreview: (event: React.MouseEvent) => void;
   onDuplicate: (event: React.MouseEvent) => void;
   onEdit: (event: React.MouseEvent) => void;
@@ -775,11 +778,11 @@ function StructureCard({
     <div
       role="button"
       tabIndex={0}
-      onClick={onOpen}
+      onClick={onSelect}
       onKeyDown={(event) => {
         if (event.key === "Enter" || event.key === " ") {
           event.preventDefault();
-          onOpen();
+          onSelect();
         }
       }}
       className={`group relative flex cursor-pointer select-none flex-col justify-between rounded-2xl border p-5 text-left transition-all duration-200 outline-none focus-visible:ring-2 focus-visible:ring-[#F5222D]/30 ${

frontend-web/src/pages/official-multi-agent-tasks/MultiAgentTasksPage.tsx

@@ -0,0 +1,584 @@
+import {
+  CalendarClock,
+  CircleAlert,
+  CircleCheck,
+  Clock3,
+  History,
+  LoaderCircle,
+  Network,
+  Pencil,
+  Pause,
+  Play,
+  Plus,
+  Search,
+  Trash2,
+  Zap,
+} from "lucide-react";
+import { useCallback, useMemo, useState } from "react";
+import { useNavigate } from "react-router-dom";
+import { Button, DialogPlugin, Drawer, Input, Popconfirm, Switch, Tabs, Tag } from "tdesign-react";
+import { toast } from "sonner";
+
+import { OFFICIAL_DEERFLOW_ROUTE_PREFIX } from "@/core/official-deerflow/routes";
+
+import AgentFlow from "./components/AgentFlow";
+import { useMultiAgentTaskActions, useMultiAgentTasks } from "./store";
+import type { MatRunView, MatTaskView } from "./types";
+import {
+  describeCron,
+  expandStages,
+  formatDateTime,
+  formatDuration,
+  formatRelative,
+  runStatusLabel,
+  taskBadgeOf,
+} from "./utils";
+import "./multi-agent-tasks.css";
+
+type StatusFilter = "all" | "running" | "enabled" | "paused";
+
+function isTaskLive(task: MatTaskView) {
+  return task.lastRun?.status === "running" || task.lastRun?.status === "pending";
+}
+
+function matchFilter(task: MatTaskView, filter: StatusFilter) {
+  if (filter === "all") return true;
+  if (filter === "running") return isTaskLive(task);
+  if (filter === "paused") return !task.enabled;
+  return task.enabled && !isTaskLive(task);
+}
+
+function runTagTheme(status: MatRunView["status"]): "success" | "danger" | "warning" | "default" {
+  if (status === "succeeded") return "success";
+  if (status === "failed") return "danger";
+  if (status === "running") return "warning";
+  return "default";
+}
+
+export default function MultiAgentTasksPage({
+  routeTaskId,
+}: {
+  routeTaskId: string | null;
+}) {
+  const navigate = useNavigate();
+  const { tasks, isLoading } = useMultiAgentTasks();
+  const actions = useMultiAgentTaskActions();
+  const [search, setSearch] = useState("");
+  const [filter, setFilter] = useState<StatusFilter>("all");
+  const [busyTaskId, setBusyTaskId] = useState<string | null>(null);
+  const [activeRunId, setActiveRunId] = useState<string | null>(null);
+
+  const selectedTask = tasks.find((task) => task.id === routeTaskId) ?? null;
+
+  const openTask = useCallback(
+    (taskId: string) => {
+      setActiveRunId(null);
+      navigate(
+        `${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/multi-agent-tasks/${encodeURIComponent(taskId)}`,
+      );
+    },
+    [navigate],
+  );
+
+  const closeTask = useCallback(() => {
+    setActiveRunId(null);
+    navigate(`${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/multi-agent-tasks`);
+  }, [navigate]);
+
+  const visibleTasks = useMemo(() => {
+    const keyword = search.trim().toLowerCase();
+    return tasks.filter((task) => {
+      if (!matchFilter(task, filter)) return false;
+      if (!keyword) return true;
+      return (
+        task.name.toLowerCase().includes(keyword) ||
+        (task.description ?? "").toLowerCase().includes(keyword) ||
+        task.seats.some((seat) => seat.name.toLowerCase().includes(keyword))
+      );
+    });
+  }, [tasks, search, filter]);
+
+  const stats = useMemo(() => {
+    const total = tasks.length;
+    const enabled = tasks.filter((task) => task.enabled).length;
+    const running = tasks.filter(isTaskLive).length;
+    const runTotal = tasks.reduce((sum, task) => sum + task.runs.length, 0);
+    return { total, enabled, running, runTotal };
+  }, [tasks]);
+
+  const openCreate = () => {
+    navigate(`${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/multi-agent-tasks/new`);
+  };
+
+  const openEdit = (task: MatTaskView) => {
+    navigate(
+      `${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/multi-agent-tasks/${encodeURIComponent(task.id)}/edit`,
+    );
+  };
+
+  const runTrigger = async (task: MatTaskView) => {
+    setBusyTaskId(task.id);
+    try {
+      await actions.trigger.mutateAsync(task.id);
+      setActiveRunId(null);
+      openTask(task.id);
+      toast.success("已触发执行");
+    } catch {
+      toast.error("触发失败，请重试");
+    } finally {
+      setBusyTaskId(null);
+    }
+  };
+
+  const runToggle = async (task: MatTaskView) => {
+    setBusyTaskId(task.id);
+    try {
+      await actions.setEnabled.mutateAsync({ id: task.id, enabled: !task.enabled });
+      toast.success(task.enabled ? "任务已暂停" : "任务已恢复");
+    } catch {
+      toast.error("操作失败，请重试");
+    } finally {
+      setBusyTaskId(null);
+    }
+  };
+
+  const confirmDelete = (task: MatTaskView) => {
+    const dialog = DialogPlugin.confirm({
+      header: "删除定时任务",
+      body: `确定删除「${task.name}」吗？其编排配置与运行记录将一并清除，且不可恢复。`,
+      confirmBtn: { content: "删除", theme: "danger" },
+      cancelBtn: "取消",
+      onConfirm: () => {
+        dialog.destroy();
+        void actions.remove.mutateAsync(task.id).then(() => {
+          if (selectedTask?.id === task.id) closeTask();
+          toast.success("任务已删除");
+        });
+      },
+      onClose: () => dialog.destroy(),
+    });
+  };
+
+  const selectedStages = selectedTask ? expandStages(selectedTask) : [];
+  const selectedRuns = selectedTask?.runs ?? [];
+  const activeRun =
+    selectedRuns.find((run) => run.id === activeRunId) ?? selectedRuns[0] ?? null;
+
+  return (
+    <section className="mat-page">
+      <header className="mat-page-head">
+        <div className="mat-page-head-copy">
+          <h1>多智能体定时任务</h1>
+          <p>
+            把多个智能体编排成协作编队：同阶段并行、阶段间串行，按计划自动执行并沉淀运行记录。
+          </p>
+        </div>
+        <Button theme="primary" onClick={openCreate}>
+          <Plus aria-hidden="true" size={15} />
+          新建编排任务
+        </Button>
+      </header>
+
+      <div className="mat-stats" aria-label="任务统计">
+        <div className="mat-stat">
+          <span className="mat-stat-icon is-primary" aria-hidden="true">
+            <Network size={18} />
+          </span>
+          <div className="mat-stat-copy">
+            <strong>{stats.total}</strong>
+            <span>编排任务</span>
+          </div>
+        </div>
+        <div className="mat-stat">
+          <span className="mat-stat-icon is-success" aria-hidden="true">
+            <CircleCheck size={18} />
+          </span>
+          <div className="mat-stat-copy">
+            <strong>{stats.enabled}</strong>
+            <span>已启用</span>
+          </div>
+        </div>
+        <div className="mat-stat">
+          <span className={`mat-stat-icon${stats.running ? " is-live" : " is-muted"}`} aria-hidden="true">
+            {stats.running ? <LoaderCircle size={18} className="mat-spin" /> : <Clock3 size={18} />}
+          </span>
+          <div className="mat-stat-copy">
+            <strong className={stats.running ? "is-live" : undefined}>{stats.running}</strong>
+            <span>执行中</span>
+          </div>
+        </div>
+        <div className="mat-stat">
+          <span className="mat-stat-icon is-info" aria-hidden="true">
+            <History size={18} />
+          </span>
+          <div className="mat-stat-copy">
+            <strong>{stats.runTotal}</strong>
+            <span>累计执行</span>
+          </div>
+        </div>
+      </div>
+
+      <div className="mat-body">
+        <div className="mat-toolbar">
+          <div className="mat-toolbar-search">
+            <Input
+              clearable
+              placeholder="搜索任务名称 / 描述 / 智能体"
+              prefixIcon={<Search aria-hidden="true" size={15} />}
+              value={search}
+              onChange={(value) => setSearch(String(value ?? ""))}
+            />
+          </div>
+          <Tabs
+            className="mat-toolbar-tabs"
+            value={filter}
+            onChange={(value) => setFilter(value as StatusFilter)}
+          >
+            <Tabs.TabPanel label="全部" value="all" />
+            <Tabs.TabPanel label="执行中" value="running" />
+            <Tabs.TabPanel label="已启用" value="enabled" />
+            <Tabs.TabPanel label="已暂停" value="paused" />
+          </Tabs>
+        </div>
+
+        {isLoading ? (
+          <div className="mat-state">
+            <Clock3 aria-hidden="true" className="mat-spin" size={18} />
+            正在加载编排任务…
+          </div>
+        ) : visibleTasks.length ? (
+          <div className="mat-grid">
+            {visibleTasks.map((task) => {
+              const badge = taskBadgeOf(task);
+              const live = isTaskLive(task);
+              const busy = busyTaskId === task.id;
+              return (
+                <article className="mat-card" key={task.id}>
+                  <header className="mat-card-head" onClick={() => openTask(task.id)}>
+                    <span className="mat-card-glyph" aria-hidden="true">
+                      <Network size={17} />
+                    </span>
+                    <div className="mat-card-title">
+                      <h3>{task.name}</h3>
+                      <p>
+                        <CalendarClock aria-hidden="true" size={13} />
+                        {describeCron(task.cron_expr)}
+                      </p>
+                    </div>
+                    <Tag size="small" theme={badge.theme} variant="light">
+                      {badge.label}
+                    </Tag>
+                  </header>
+                  {task.description && (
+                    <p className="mat-card-desc" onClick={() => openTask(task.id)}>
+                      {task.description}
+                    </p>
+                  )}
+                  <div className="mat-card-flow" onClick={() => openTask(task.id)}>
+                    <AgentFlow
+                      seats={task.seats}
+                      size="sm"
+                      stages={expandStages(task)}
+                    />
+                  </div>
+                  <footer className="mat-card-foot">
+                    <div className="mat-card-meta">
+                      <span title={task.nextRunAt ?? undefined}>
+                        下次 {task.enabled ? formatDateTime(task.nextRunAt) : "已暂停"}
+                      </span>
+                      <span>
+                        上次 {task.lastRun ? formatRelative(task.lastRun.started_at) : "未执行"}
+                        {task.lastRun && task.lastRun.status !== "pending" && (
+                          <em className={`mat-run-text is-${task.lastRun.status}`}>
+                            · {runStatusLabel(task.lastRun.status)}
+                          </em>
+                        )}
+                      </span>
+                    </div>
+                    <div className="mat-card-ops">
+                      <Button
+                        disabled={busy}
+                        loading={busy}
+                        size="small"
+                        theme="primary"
+                        variant="text"
+                        onClick={() => void runTrigger(task)}
+                      >
+                        <Zap aria-hidden="true" size={14} />
+                        立即执行
+                      </Button>
+                      <Button
+                        disabled={busy}
+                        size="small"
+                        variant="text"
+                        onClick={() => void runToggle(task)}
+                      >
+                        {task.enabled ? (
+                          <Pause aria-hidden="true" size={14} />
+                        ) : (
+                          <Play aria-hidden="true" size={14} />
+                        )}
+                        {task.enabled ? "暂停" : "恢复"}
+                      </Button>
+                      <Button size="small" variant="text" onClick={() => openEdit(task)}>
+                        <Pencil aria-hidden="true" size={14} />
+                        编辑
+                      </Button>
+                      <Popconfirm
+                        content="删除后不可恢复，确定删除该任务？"
+                        placement="bottom-right"
+                        theme="danger"
+                        onConfirm={() => confirmDelete(task)}
+                      >
+                        <Button className="mat-danger-btn" size="small" variant="text">
+                          <Trash2 aria-hidden="true" size={14} />
+                        </Button>
+                      </Popconfirm>
+                    </div>
+                  </footer>
+                </article>
+              );
+            })}
+          </div>
+        ) : (
+          <div className="mat-state mat-state--empty">
+            <span className="mat-state-glyph" aria-hidden="true">
+              <Network size={22} />
+            </span>
+            <h3>{tasks.length ? "没有符合条件的任务" : "还没有多智能体定时任务"}</h3>
+            <p>
+              {tasks.length
+                ? "换个关键词或筛选条件试试。"
+                : "创建一个编排任务，把多个智能体按并行 / 串行组织起来定时执行。"}
+            </p>
+            {!tasks.length && (
+              <Button theme="primary" onClick={openCreate}>
+                <Plus aria-hidden="true" size={15} />
+                新建编排任务
+              </Button>
+            )}
+          </div>
+        )}
+
+        <p className="mat-footnote">
+          <CircleAlert aria-hidden="true" size={13} />
+          前端演示阶段：任务数据保存在本浏览器，执行过程为本地模拟；后端服务接入后自动切换。
+        </p>
+      </div>
+
+      <Drawer
+        className="mat-detail-drawer"
+        closeBtn
+        destroyOnClose
+        footer={
+          selectedTask ? (
+            <div className="mat-detail-footer">
+              <Button variant="outline" onClick={() => openEdit(selectedTask)}>
+                <Pencil aria-hidden="true" size={15} />
+                编辑编排
+              </Button>
+              <Button
+                variant="outline"
+                onClick={() => void runToggle(selectedTask)}
+              >
+                {selectedTask.enabled ? (
+                  <Pause aria-hidden="true" size={15} />
+                ) : (
+                  <Play aria-hidden="true" size={15} />
+                )}
+                {selectedTask.enabled ? "暂停任务" : "恢复任务"}
+              </Button>
+              <Button
+                className="mat-danger-btn"
+                theme="danger"
+                variant="outline"
+                onClick={() => confirmDelete(selectedTask)}
+              >
+                <Trash2 aria-hidden="true" size={15} />
+                删除
+              </Button>
+              <Button
+                disabled={busyTaskId === selectedTask.id}
+                loading={busyTaskId === selectedTask.id}
+                theme="primary"
+                onClick={() => void runTrigger(selectedTask)}
+              >
+                <Zap aria-hidden="true" size={15} />
+                立即执行
+              </Button>
+            </div>
+          ) : null
+        }
+        header={
+          selectedTask ? (
+            <div className="mat-detail-header">
+              <span className="mat-detail-glyph" aria-hidden="true">
+                <Network size={17} />
+              </span>
+              <div className="mat-detail-heading">
+                <h3>{selectedTask.name}</h3>
+                <p>{describeCron(selectedTask.cron_expr)}</p>
+              </div>
+              {(() => {
+                const badge = taskBadgeOf(selectedTask);
+                return (
+                  <Tag theme={badge.theme} variant="light">
+                    {badge.label}
+                  </Tag>
+                );
+              })()}
+              <Switch
+                value={selectedTask.enabled}
+                onChange={() => void runToggle(selectedTask)}
+              />
+            </div>
+          ) : null
+        }
+        placement="right"
+        size="680px"
+        visible={!!selectedTask}
+        onClose={closeTask}
+      >
+        {selectedTask && (
+          <div className="mat-detail">
+            {selectedTask.description && <p className="mat-detail-desc">{selectedTask.description}</p>}
+
+            <div className="mat-detail-grid">
+              <div>
+                <span>下次执行</span>
+                <strong>
+                  {selectedTask.enabled ? formatDateTime(selectedTask.nextRunAt) : "已暂停"}
+                </strong>
+              </div>
+              <div>
+                <span>最近执行</span>
+                <strong>
+                  {selectedTask.lastRun
+                    ? `${formatRelative(selectedTask.lastRun.started_at)} · ${runStatusLabel(selectedTask.lastRun.status)}`
+                    : "未执行"}
+                </strong>
+              </div>
+              <div>
+                <span>累计执行</span>
+                <strong>{selectedTask.runs.length} 次</strong>
+              </div>
+              <div>
+                <span>智能体 / 阶段</span>
+                <strong>
+                  {selectedTask.seats.length} 个 · {selectedStages.length} 阶段
+                </strong>
+              </div>
+            </div>
+
+            <section className="mat-detail-section">
+              <header className="mat-detail-section-head">
+                <h4>协作编排</h4>
+                {activeRun && (activeRun.status === "running" || activeRun.status === "pending") && (
+                  <span className="mat-run-text is-running">
+                    阶段 {Math.min(activeRun.stage_index + 1, activeRun.stage_total)} /{" "}
+                    {activeRun.stage_total} 执行中…
+                  </span>
+                )}
+              </header>
+              <div className="mat-detail-flow">
+                <AgentFlow
+                  activeStageIndex={
+                    activeRun && activeRun.status !== "succeeded" ? activeRun.stage_index : null
+                  }
+                  seatStates={activeRun?.seat_states ?? null}
+                  seats={selectedTask.seats}
+                  showGoals
+                  size="sm"
+                  stageGoals={selectedTask.stage_goals}
+                  stages={selectedStages}
+                />
+              </div>
+              {activeRun && (
+                <div className="mat-flow-legend">
+                  <span>
+                    <i className="mat-seat-dot is-done" /> 完成
+                  </span>
+                  <span>
+                    <i className="mat-seat-dot is-running" /> 执行中
+                  </span>
+                  <span>
+                    <i className="mat-seat-dot is-waiting" /> 等待
+                  </span>
+                  <span>
+                    <i className="mat-seat-dot is-failed" /> 失败
+                  </span>
+                </div>
+              )}
+            </section>
+
+            <section className="mat-detail-section">
+              <header className="mat-detail-section-head">
+                <h4>任务指令</h4>
+              </header>
+              <p className="mat-detail-prompt">{selectedTask.prompt}</p>
+              {selectedTask.coordinator_prompt && (
+                <>
+                  <header className="mat-detail-section-head mat-detail-section-head--sub">
+                    <h5>总控提示</h5>
+                  </header>
+                  <p className="mat-detail-prompt is-sub">{selectedTask.coordinator_prompt}</p>
+                </>
+              )}
+            </section>
+
+            <section className="mat-detail-section">
+              <header className="mat-detail-section-head">
+                <h4>运行记录</h4>
+                {!!selectedRuns.length && (
+                  <small>点击记录可在上方流程图回看各智能体状态</small>
+                )}
+              </header>
+              {selectedRuns.length ? (
+                <div className="mat-run-list">
+                  {selectedRuns.map((run) => {
+                    const active = activeRun?.id === run.id;
+                    const duration = run.finished_at
+                      ? formatDuration(
+                          new Date(run.finished_at).getTime() -
+                            new Date(run.started_at).getTime(),
+                        )
+                      : null;
+                    return (
+                      <button
+                        className={`mat-run-item${active ? " is-active" : ""}`}
+                        key={run.id}
+                        onClick={() => setActiveRunId(run.id)}
+                        type="button"
+                      >
+                        <Tag size="small" theme={runTagTheme(run.status)} variant="light">
+                          {runStatusLabel(run.status)}
+                        </Tag>
+                        <span className="mat-run-item-time">
+                          {formatDateTime(run.started_at)}
+                        </span>
+                        <span className="mat-run-item-trigger">
+                          {run.trigger === "manual" ? "手动触发" : "按计划执行"}
+                        </span>
+                        <span className="mat-run-item-duration">
+                          {duration ?? `阶段 ${Math.min(run.stage_index + 1, run.stage_total)}/${run.stage_total}`}
+                        </span>
+                      </button>
+                    );
+                  })}
+                </div>
+              ) : (
+                <p className="mat-detail-empty">还没有执行记录，可点「立即执行」体验完整流程。</p>
+              )}
+              {activeRun?.error && (
+                <p className="mat-detail-error">
+                  <CircleAlert aria-hidden="true" size={14} />
+                  {activeRun.error}
+                </p>
+              )}
+            </section>
+          </div>
+        )}
+      </Drawer>
+    </section>
+  );
+}

frontend-web/src/pages/official-multi-agent-tasks/TaskEditPage.tsx

@@ -0,0 +1,942 @@
+/**
+ * 多智能体定时任务 · 独立编辑页。
+ *
+ * 布局：顶部是紧凑表单条（任务名称 + 任务描述），其余表单项（执行频率、任务指令、
+ * 总控提示、启用开关）收进「执行设置」弹窗；主体是智能体编排。
+ * 编排交互与业务链条编辑器一致（复制其 @dnd-kit 模式 + 复用 chain-stages 纯函数）：
+ * - 左栏候选池：拖左侧手柄到右侧任意阶段 / 新建阶段落区；点卡片本体 = 追加为新阶段。
+ * - 右栏阶段组：阶段是 droppable，席位是 sortable；席位可跨阶段拖动（变并行）或拖到
+ *   「新建阶段」落区独立成阶段；按钮兜底（并入上/下一阶段、独立成阶段、移除）。
+ * 保存时弹出确认弹窗：任务摘要 + 编排流程图（同阶段并行 · 阶段间串行）+ 图例。
+ */
+
+import {
+  ArrowLeft,
+  Bot,
+  ChevronDown,
+  ChevronUp,
+  Eye,
+  GripVertical,
+  Plus,
+  Save,
+  Search,
+  Settings2,
+  SplitSquareVertical,
+  Trash2,
+  X,
+} from "lucide-react";
+import { useCallback, useEffect, useMemo, useState } from "react";
+import { useNavigate } from "react-router-dom";
+import {
+  DndContext,
+  DragOverlay,
+  PointerSensor,
+  closestCenter,
+  useDraggable,
+  useDroppable,
+  useSensor,
+  useSensors,
+  type DragEndEvent,
+  type DragStartEvent,
+} from "@dnd-kit/core";
+import {
+  SortableContext,
+  useSortable,
+  verticalListSortingStrategy,
+} from "@dnd-kit/sortable";
+import { CSS } from "@dnd-kit/utilities";
+import { Button, Dialog, Input, Switch, Textarea } from "tdesign-react";
+import { toast } from "sonner";
+
+import { useAgents } from "@/core/agents";
+import type { Agent } from "@/core/agents";
+import {
+  type ScheduleConfig,
+  buildCron,
+  parseCron,
+} from "@/components/workspace/scheduled-tasks/schedule-builder";
+import {
+  addEmptyStage,
+  addSeatAsNewStage,
+  addSeatToStage,
+  countSeats,
+  deleteStage,
+  findStageIndexOfSeat,
+  moveSeatToStage,
+  moveStage,
+  removeSeat,
+  setStageGoal,
+  splitSeatToNewStage,
+  stageGroupsToStageGoals,
+  stageGroupsToStagePayload,
+  type StageGroups,
+} from "@/roundtable-planning/lib/chain-stages";
+
+import { OFFICIAL_DEERFLOW_ROUTE_PREFIX } from "@/core/official-deerflow/routes";
+
+import AgentFlow from "./components/AgentFlow";
+import AgentDetailDrawer, {
+  type AgentDetailTarget,
+} from "./components/AgentDetailDrawer";
+import SchedulePicker from "./components/SchedulePicker";
+import { useMultiAgentTaskActions, useMultiAgentTasks } from "./store";
+import type { MatSeat, MatTask, MatTaskPayload } from "./types";
+import {
+  defaultScheduleConfig,
+  describeCron,
+  describeScheduleConfig,
+  expandStages,
+} from "./utils";
+
+const POOL_PREFIX = "matpool::";
+const SEAT_PREFIX = "matseat::";
+const STAGE_PREFIX = "matstage::";
+const NEWSTAGE_DROPPABLE_ID = "mat-newstage";
+
+const MAX_SEATS = 20;
+const MIN_SEATS = 2;
+const DEFAULT_TIMEZONE = "Asia/Shanghai";
+
+type ActiveDrag = { name: string } | null;
+
+function seatFromAgent(agent: Agent): MatSeat {
+  return {
+    agent_id: agent.id,
+    name: agent.name,
+    description: agent.description ?? undefined,
+  };
+}
+
+function draftFromTask(task: MatTask) {
+  const stages = expandStages(task);
+  const seatById = new Map(task.seats.map((seat) => [seat.agent_id, seat]));
+  return {
+    name: task.name,
+    description: task.description ?? "",
+    schedule: parseCron(task.cron_expr) ?? defaultScheduleConfig(),
+    enabled: task.enabled,
+    prompt: task.prompt,
+    coordinatorPrompt: task.coordinator_prompt ?? "",
+    stageGroups: stages.map((stage, index) => ({
+      seats: stage
+        .map((agentId) => seatById.get(agentId))
+        .filter((seat): seat is MatSeat => !!seat),
+      goal: task.stage_goals?.[index] ?? "",
+    })) as StageGroups,
+  };
+}
+
+function groupsToPayloadSeats(groups: StageGroups): MatSeat[] {
+  return groups
+    .filter((group) => group.seats.length > 0)
+    .flatMap((group) => group.seats as MatSeat[]);
+}
+
+// ── 左栏：候选智能体卡片（点卡片看详情 / 加号或手柄加入）──────────────────
+
+function PoolAgentCard({
+  agent,
+  used,
+  onAdd,
+  onShowDetail,
+}: {
+  agent: Agent;
+  used: boolean;
+  onAdd: () => void;
+  onShowDetail: () => void;
+}) {
+  const { attributes, listeners, setNodeRef, isDragging } = useDraggable({
+    id: POOL_PREFIX + agent.id,
+    disabled: used,
+  });
+  return (
+    <div
+      className={`mat-pool-item${used ? " is-used" : ""}${isDragging ? " is-dragging" : ""}`}
+      ref={setNodeRef}
+      {...attributes}
+    >
+      <button
+        aria-label={used ? "已在编排中" : "拖动加入编排"}
+        className="mat-pool-grip"
+        onClick={(event) => event.stopPropagation()}
+        title={used ? "已在编排中" : "拖动到右侧阶段加入（可与其它智能体并行）"}
+        type="button"
+        {...(used ? {} : listeners)}
+      >
+        <GripVertical size={14} />
+      </button>
+      <button className="mat-pool-main" onClick={onShowDetail} type="button">
+        <span className="mat-pool-icon" aria-hidden="true">
+          <Bot size={13} />
+        </span>
+        <span className="mat-pool-copy">
+          <span className="mat-pool-item-name">{agent.name}</span>
+          <span className="mat-pool-item-desc">{agent.description || "暂无简介"}</span>
+        </span>
+      </button>
+      {used ? (
+        <span className="mat-pool-used">已加入</span>
+      ) : (
+        <button
+          aria-label={`加入 ${agent.name}`}
+          className="mat-pool-join"
+          onClick={(event) => {
+            event.stopPropagation();
+            onAdd();
+          }}
+          title="追加为新阶段"
+          type="button"
+        >
+          <Plus size={13} />
+        </button>
+      )}
+    </div>
+  );
+}
+
+// ── 右栏：单个席位长条（sortable）─────────────────────────────────────────
+
+function SeatStrip({
+  seat,
+  globalIndex,
+  isFirstStage,
+  isLastStage,
+  onRemove,
+  onMergeUp,
+  onMergeDown,
+  onSplit,
+  onDetail,
+}: {
+  seat: MatSeat;
+  globalIndex: number;
+  isFirstStage: boolean;
+  isLastStage: boolean;
+  onRemove: () => void;
+  onMergeUp: () => void;
+  onMergeDown: () => void;
+  onSplit: () => void;
+  onDetail: () => void;
+}) {
+  const { attributes, listeners, setNodeRef, transform, transition, isDragging } =
+    useSortable({ id: SEAT_PREFIX + seat.agent_id });
+  return (
+    <div
+      className={`mat-seat-row${isDragging ? " is-dragging" : ""}`}
+      ref={setNodeRef}
+      style={{ transform: CSS.Transform.toString(transform), transition }}
+    >
+      <button
+        aria-label="拖动调整编排"
+        className="mat-seat-row-grip"
+        title="拖动到其它阶段（并行）/ 排序 / 拖到底部虚线框独立成阶段"
+        type="button"
+        {...attributes}
+        {...listeners}
+      >
+        <GripVertical size={14} />
+      </button>
+      <span className="mat-seat-row-no">{globalIndex + 1}</span>
+      <span className="mat-seat-row-icon" aria-hidden="true">
+        <Bot size={13} />
+      </span>
+      <div className="mat-seat-row-copy" title={seat.description}>
+        <span className="mat-seat-row-name">{seat.name}</span>
+        {seat.description && <span className="mat-seat-row-desc">{seat.description}</span>}
+      </div>
+      <span className="mat-seat-row-ops">
+        <button
+          aria-label={`查看 ${seat.name} 详情`}
+          onClick={onDetail}
+          title="查看详情"
+          type="button"
+        >
+          <Eye size={14} />
+        </button>
+        <button
+          aria-label="并入上一阶段（与其并行）"
+          disabled={isFirstStage}
+          onClick={onMergeUp}
+          title="并入上一阶段（与其并行）"
+          type="button"
+        >
+          <ChevronUp size={14} />
+        </button>
+        <button
+          aria-label="独立成阶段"
+          onClick={onSplit}
+          title="独立成阶段"
+          type="button"
+        >
+          <SplitSquareVertical size={14} />
+        </button>
+        <button
+          aria-label="并入下一阶段（与其并行）"
+          disabled={isLastStage}
+          onClick={onMergeDown}
+          title="并入下一阶段（与其并行）"
+          type="button"
+        >
+          <ChevronDown size={14} />
+        </button>
+        <button aria-label="移除智能体" onClick={onRemove} title="移除" type="button">
+          <X size={14} />
+        </button>
+      </span>
+    </div>
+  );
+}
+
+// ── 右栏：阶段容器（droppable + 组内 sortable）────────────────────────────
+
+function StageBox({
+  group,
+  stageIdx,
+  stageCount,
+  startIndex,
+  onGoalChange,
+  onStageReorder,
+  onStageDelete,
+  onRemoveSeat,
+  onMergeSeat,
+  onSplitSeat,
+  onSeatDetail,
+}: {
+  group: StageGroups[number];
+  stageIdx: number;
+  stageCount: number;
+  startIndex: number;
+  onGoalChange: (goal: string) => void;
+  onStageReorder: (dir: -1 | 1) => void;
+  onStageDelete: () => void;
+  onRemoveSeat: (agentId: string) => void;
+  onMergeSeat: (agentId: string, targetStage: number) => void;
+  onSplitSeat: (agentId: string) => void;
+  onSeatDetail: (seat: MatSeat) => void;
+}) {
+  const { setNodeRef, isOver } = useDroppable({ id: STAGE_PREFIX + stageIdx });
+  const seats = group.seats as MatSeat[];
+  const isParallel = seats.length > 1;
+  const isFirstStage = stageIdx === 0;
+  const isLastStage = stageIdx === stageCount - 1;
+  const seatItemIds = seats.map((seat) => SEAT_PREFIX + seat.agent_id);
+
+  return (
+    <article className={`mat-stage-box${isOver ? " is-over" : ""}`}>
+      <header className="mat-stage-box-head">
+        <span className="mat-stage-box-no">阶段 {stageIdx + 1}</span>
+        {isParallel ? (
+          <span className="mat-flow-mode is-parallel">并行 ×{seats.length}</span>
+        ) : (
+          <span className="mat-flow-mode is-serial">串行</span>
+        )}
+        <span className="mat-stage-box-ops">
+          <button
+            aria-label="阶段上移"
+            disabled={isFirstStage}
+            onClick={() => onStageReorder(-1)}
+            title="上移"
+            type="button"
+          >
+            <ChevronUp size={14} />
+          </button>
+          <button
+            aria-label="阶段下移"
+            disabled={isLastStage}
+            onClick={() => onStageReorder(1)}
+            title="下移"
+            type="button"
+          >
+            <ChevronDown size={14} />
+          </button>
+          <button
+            aria-label="删除阶段"
+            onClick={onStageDelete}
+            title="删除整个阶段"
+            type="button"
+          >
+            <Trash2 size={14} />
+          </button>
+        </span>
+      </header>
+      <Textarea
+        autosize
+        className="mat-stage-goal"
+        maxlength={120}
+        placeholder="本阶段目标（可选）：本阶段要产出什么、如何衔接上一阶段"
+        value={group.goal ?? ""}
+        onChange={(value) => onGoalChange(String(value ?? ""))}
+      />
+      <div className="mat-stage-seats" ref={setNodeRef}>
+        {seats.length === 0 ? (
+          <div className="mat-stage-seat-empty">空阶段 —— 把左侧智能体拖到此处，或删除该阶段。</div>
+        ) : (
+          <SortableContext items={seatItemIds} strategy={verticalListSortingStrategy}>
+            {seats.map((seat, index) => (
+              <SeatStrip
+                globalIndex={startIndex + index}
+                isLastStage={isLastStage}
+                isFirstStage={isFirstStage}
+                key={seat.agent_id}
+                onDetail={() => onSeatDetail(seat)}
+                onMergeDown={() => onMergeSeat(seat.agent_id, stageIdx + 1)}
+                onMergeUp={() => onMergeSeat(seat.agent_id, stageIdx - 1)}
+                onRemove={() => onRemoveSeat(seat.agent_id)}
+                onSplit={() => onSplitSeat(seat.agent_id)}
+                seat={seat}
+              />
+            ))}
+          </SortableContext>
+        )}
+      </div>
+    </article>
+  );
+}
+
+// ── 「新建阶段」落区（点击新增 / 拖席位到此独立成最后阶段）────────────────
+
+function NewStageDropZone({
+  hasEmptyStage,
+  onAddStage,
+}: {
+  hasEmptyStage: boolean;
+  onAddStage: () => void;
+}) {
+  const { setNodeRef, isOver } = useDroppable({ id: NEWSTAGE_DROPPABLE_ID });
+  return (
+    <button
+      className={`mat-stage-dropzone${isOver ? " is-over" : ""}`}
+      disabled={hasEmptyStage}
+      onClick={onAddStage}
+      ref={setNodeRef}
+      title={hasEmptyStage ? "已有空阶段，先填充它" : "新建一个串在最后的阶段"}
+      type="button"
+    >
+      <Plus size={15} />
+      新建阶段（串在最后） · 或把席位拖到此处独立成阶段
+    </button>
+  );
+}
+
+// ── 页面 ──────────────────────────────────────────────────────────────────
+
+export default function TaskEditPage({
+  mode,
+  taskId,
+}: {
+  mode: "create" | "edit";
+  taskId: string | null;
+}) {
+  const navigate = useNavigate();
+  const { tasks, isLoading: tasksLoading } = useMultiAgentTasks();
+  const actions = useMultiAgentTaskActions();
+  const [poolSearch, setPoolSearch] = useState("");
+  const { agents, isLoading: agentsLoading } = useAgents({
+    search: poolSearch.trim() || undefined,
+  });
+
+  const [name, setName] = useState("");
+  const [description, setDescription] = useState("");
+  const [schedule, setSchedule] = useState<ScheduleConfig>(defaultScheduleConfig());
+  const [enabled, setEnabled] = useState(true);
+  const [prompt, setPrompt] = useState("");
+  const [coordinatorPrompt, setCoordinatorPrompt] = useState("");
+  const [stageGroups, setStageGroups] = useState<StageGroups>([]);
+  const [initialized, setInitialized] = useState(false);
+
+  const [settingsOpen, setSettingsOpen] = useState(false);
+  const [confirmOpen, setConfirmOpen] = useState(false);
+  const [saving, setSaving] = useState(false);
+  const [activeDrag, setActiveDrag] = useState<ActiveDrag>(null);
+  const [detailTarget, setDetailTarget] = useState<AgentDetailTarget | null>(null);
+
+  const editingTask = useMemo(
+    () => (mode === "edit" ? (tasks.find((task) => task.id === taskId) ?? null) : null),
+    [mode, taskId, tasks],
+  );
+
+  useEffect(() => {
+    if (initialized || mode === "create") {
+      setInitialized(true);
+      return;
+    }
+    if (!tasksLoading && taskId) {
+      if (editingTask) {
+        const draft = draftFromTask(editingTask);
+        setName(draft.name);
+        setDescription(draft.description);
+        setSchedule(draft.schedule);
+        setEnabled(draft.enabled);
+        setPrompt(draft.prompt);
+        setCoordinatorPrompt(draft.coordinatorPrompt);
+        setStageGroups(draft.stageGroups);
+      }
+      setInitialized(true);
+    }
+  }, [editingTask, initialized, mode, taskId, tasksLoading]);
+
+  const sensors = useSensors(
+    useSensor(PointerSensor, { activationConstraint: { distance: 5 } }),
+  );
+
+  const usedAgentIds = useMemo(
+    () => new Set(stageGroups.flatMap((group) => group.seats.map((seat) => seat.agent_id))),
+    [stageGroups],
+  );
+  const seatCount = useMemo(() => countSeats(stageGroups), [stageGroups]);
+  const hasEmptyStage = stageGroups.some((group) => group.seats.length === 0);
+
+  const backToList = useCallback(() => {
+    navigate(`${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/multi-agent-tasks`);
+  }, [navigate]);
+
+  // ── 编排操作（全部复用 chain-stages 纯函数，agent_id 定位）───────────────
+  const addAgentAsStage = useCallback((agent: Agent) => {
+    if (seatCount >= MAX_SEATS) {
+      toast.warning(`编排最多 ${MAX_SEATS} 个智能体`);
+      return;
+    }
+    setStageGroups((prev) => addSeatAsNewStage(prev, seatFromAgent(agent)));
+  }, [seatCount]);
+
+  const onDragStart = useCallback((event: DragStartEvent) => {
+    const id = String(event.active.id);
+    const agentId = id.startsWith(POOL_PREFIX)
+      ? id.slice(POOL_PREFIX.length)
+      : id.startsWith(SEAT_PREFIX)
+        ? id.slice(SEAT_PREFIX.length)
+        : null;
+    if (!agentId) return;
+    const seatName =
+      agents.find((agent) => agent.id === agentId)?.name ??
+      stageGroups
+        .flatMap((group) => group.seats as MatSeat[])
+        .find((seat) => seat.agent_id === agentId)?.name ??
+      agentId;
+    setActiveDrag({ name: seatName });
+  }, [agents, stageGroups]);
+
+  const onDragEnd = useCallback(
+    (event: DragEndEvent) => {
+      setActiveDrag(null);
+      const activeId = String(event.active.id);
+      const overId = event.over ? String(event.over.id) : null;
+      if (!overId) return;
+
+      let targetStage = -1;
+      let atIndex: number | undefined;
+      if (overId === NEWSTAGE_DROPPABLE_ID) {
+        targetStage = -1;
+      } else if (overId.startsWith(STAGE_PREFIX)) {
+        targetStage = Number(overId.slice(STAGE_PREFIX.length));
+      } else if (overId.startsWith(SEAT_PREFIX)) {
+        const overAgent = overId.slice(SEAT_PREFIX.length);
+        targetStage = findStageIndexOfSeat(stageGroups, overAgent);
+        atIndex = stageGroups[targetStage]?.seats.findIndex(
+          (seat) => seat.agent_id === overAgent,
+        );
+      }
+
+      const toNewStage = overId === NEWSTAGE_DROPPABLE_ID;
+
+      if (activeId.startsWith(POOL_PREFIX)) {
+        const agentId = activeId.slice(POOL_PREFIX.length);
+        if (usedAgentIds.has(agentId)) return;
+        const agent = agents.find((candidate) => candidate.id === agentId);
+        if (!agent || seatCount >= MAX_SEATS) return;
+        const seat = seatFromAgent(agent);
+        setStageGroups((prev) =>
+          toNewStage || targetStage < 0
+            ? addSeatAsNewStage(prev, seat)
+            : addSeatToStage(prev, seat, targetStage, atIndex),
+        );
+        return;
+      }
+
+      if (activeId.startsWith(SEAT_PREFIX)) {
+        const agentId = activeId.slice(SEAT_PREFIX.length);
+        setStageGroups((prev) => {
+          if (toNewStage) {
+            const withEmpty = addEmptyStage(prev);
+            return moveSeatToStage(withEmpty, agentId, withEmpty.length - 1);
+          }
+          if (targetStage < 0) return prev;
+          return moveSeatToStage(prev, agentId, targetStage, atIndex);
+        });
+      }
+    },
+    [agents, seatCount, stageGroups, usedAgentIds],
+  );
+
+  // ── 保存：校验 → 确认弹窗 → 提交 ────────────────────────────────────────
+  const previewGroups = stageGroups.filter((group) => group.seats.length > 0);
+  const previewStages = previewGroups.map((group) =>
+    (group.seats as MatSeat[]).map((seat) => seat.agent_id),
+  );
+  const previewSeats = groupsToPayloadSeats(stageGroups);
+
+  const tryOpenConfirm = () => {
+    if (!name.trim()) {
+      toast.warning("请填写任务名称");
+      return;
+    }
+    if (seatCount < MIN_SEATS) {
+      toast.warning(`编排至少需要 ${MIN_SEATS} 个智能体`);
+      return;
+    }
+    if (hasEmptyStage) {
+      toast.warning("还有空阶段，请填充或删除");
+      return;
+    }
+    if (!prompt.trim()) {
+      setSettingsOpen(true);
+      toast.warning("请先在「执行设置」中填写任务指令");
+      return;
+    }
+    setConfirmOpen(true);
+  };
+
+  const submitTask = async () => {
+    setSaving(true);
+    try {
+      // 填了总控提示也强制落显式分层（该提示仅分层模式注入），与业务链条语义一致。
+      const forceExplicit = coordinatorPrompt.trim().length > 0;
+      const payload: MatTaskPayload = {
+        name: name.trim(),
+        description: description.trim() || undefined,
+        prompt: prompt.trim(),
+        coordinator_prompt: coordinatorPrompt.trim() || undefined,
+        cron_expr: buildCron(schedule),
+        timezone: editingTask?.timezone || DEFAULT_TIMEZONE,
+        enabled,
+        seats: previewSeats,
+        stages: stageGroupsToStagePayload(stageGroups, forceExplicit),
+        stage_goals: stageGroupsToStageGoals(stageGroups, forceExplicit),
+      };
+      if (mode === "edit" && taskId) {
+        await actions.update.mutateAsync({ id: taskId, payload });
+        toast.success("任务已更新");
+      } else {
+        await actions.create.mutateAsync(payload);
+        toast.success("任务已创建");
+      }
+      setConfirmOpen(false);
+      backToList();
+    } catch {
+      toast.error("保存失败，请重试");
+    } finally {
+      setSaving(false);
+    }
+  };
+
+  if (mode === "edit" && !tasksLoading && initialized && !editingTask) {
+    return (
+      <section className="mat-edit-page">
+        <div className="mat-state mat-state--empty">
+          <h3>任务不存在或已被删除</h3>
+          <Button theme="primary" onClick={backToList}>
+            返回任务列表
+          </Button>
+        </div>
+      </section>
+    );
+  }
+
+  let seatCursor = 0;
+  const settingsSummary = `${describeScheduleConfig(schedule)} · ${enabled ? "已启用" : "已暂停"}`;
+
+  return (
+    <section className="mat-edit-page">
+      <header className="mat-edit-header">
+        <button aria-label="返回任务列表" className="mat-edit-back" onClick={backToList} type="button">
+          <ArrowLeft size={18} />
+        </button>
+        <div className="mat-edit-heading">
+          <h1>{mode === "edit" ? `编辑：${editingTask?.name ?? "多智能体定时任务"}` : "新建多智能体定时任务"}</h1>
+          <p>同阶段并行执行 · 阶段之间串行执行；拖拽或点击左侧候选智能体加入编排</p>
+        </div>
+        <Button
+          variant="outline"
+          onClick={() => setSettingsOpen(true)}
+          title="执行频率 / 任务指令 / 总控提示 / 启用"
+        >
+          <Settings2 aria-hidden="true" size={15} />
+          执行设置
+          <span className="mat-settings-summary">{settingsSummary}</span>
+        </Button>
+        <Button theme="primary" onClick={tryOpenConfirm}>
+          <Save aria-hidden="true" size={15} />
+          保存任务
+        </Button>
+      </header>
+
+      <div className="mat-edit-formbar">
+        <label className="mat-formbar-field">
+          <span className="mat-formbar-label">
+            任务名称<i>*</i>
+          </span>
+          <Input
+            clearable
+            maxlength={40}
+            placeholder="例如：每日舆情早报"
+            value={name}
+            onChange={(value) => setName(String(value ?? ""))}
+          />
+        </label>
+        <label className="mat-formbar-field mat-formbar-field--wide">
+          <span className="mat-formbar-label">任务描述</span>
+          <Input
+            clearable
+            maxlength={200}
+            placeholder="一句话说明这个编排任务要达成什么（可选）"
+            value={description}
+            onChange={(value) => setDescription(String(value ?? ""))}
+          />
+        </label>
+      </div>
+
+      <div className="mat-edit-formbar mat-edit-formbar--area">
+        <span className="mat-formbar-label">总控提示</span>
+        <Textarea
+          autosize={{ minRows: 1, maxRows: 4 }}
+          className="mat-formbar-textarea"
+          maxlength={1000}
+          placeholder="写给总控（协调者）的编排要求，例如各阶段如何衔接、产出格式约束等（可选）"
+          value={coordinatorPrompt}
+          onChange={(value) => setCoordinatorPrompt(String(value ?? ""))}
+        />
+      </div>
+
+      <div className="mat-edit-body">
+        <section className="mat-edit-card">
+          <DndContext
+            collisionDetection={closestCenter}
+            onDragEnd={onDragEnd}
+            onDragStart={onDragStart}
+            sensors={sensors}
+          >
+            <div className="mat-edit-orchestra">
+              <aside className="mat-pool">
+                <header className="mat-pool-head">
+                  <span>候选智能体</span>
+                  <em>
+                    {seatCount}/{MAX_SEATS} 已编排 · {agents.length} 个可选
+                  </em>
+                </header>
+                <div className="mat-pool-search">
+                  <Search aria-hidden="true" size={15} />
+                  <input
+                    onChange={(event) => setPoolSearch(event.target.value)}
+                    placeholder="搜索智能体"
+                    value={poolSearch}
+                  />
+                </div>
+                <div className="mat-pool-list">
+                  {agentsLoading ? (
+                    <p className="mat-pool-tip">正在加载智能体…</p>
+                  ) : agents.length ? (
+                    agents.map((agent) => (
+                      <PoolAgentCard
+                        agent={agent}
+                        key={agent.id}
+                        onAdd={() => addAgentAsStage(agent)}
+                        onShowDetail={() =>
+                          setDetailTarget({
+                            agentId: agent.id,
+                            fallbackName: agent.name,
+                            fallbackDescription: agent.description ?? undefined,
+                          })
+                        }
+                        used={usedAgentIds.has(agent.id)}
+                      />
+                    ))
+                  ) : (
+                    <p className="mat-pool-tip">没有匹配的智能体</p>
+                  )}
+                </div>
+              </aside>
+
+              <div className="mat-stage-editor">
+                <div className="mat-stage-list">
+                  {stageGroups.map((group, stageIdx) => {
+                    const startIndex = seatCursor;
+                    seatCursor += (group.seats as MatSeat[]).length;
+                    return (
+                      <StageBox
+                        group={group}
+                        key={`stage-box-${stageIdx}`}
+                        onGoalChange={(goal) =>
+                          setStageGroups((prev) => setStageGoal(prev, stageIdx, goal))
+                        }
+                        onMergeSeat={(agentId, target) =>
+                          setStageGroups((prev) => moveSeatToStage(prev, agentId, target))
+                        }
+                        onRemoveSeat={(agentId) =>
+                          setStageGroups((prev) => removeSeat(prev, agentId))
+                        }
+                        onSeatDetail={(seat) =>
+                          setDetailTarget({
+                            agentId: seat.agent_id,
+                            fallbackName: seat.name,
+                            fallbackDescription: seat.description,
+                          })
+                        }
+                        onSplitSeat={(agentId) =>
+                          setStageGroups((prev) => splitSeatToNewStage(prev, agentId))
+                        }
+                        onStageDelete={() =>
+                          setStageGroups((prev) => deleteStage(prev, stageIdx))
+                        }
+                        onStageReorder={(dir) =>
+                          setStageGroups((prev) => moveStage(prev, stageIdx, dir))
+                        }
+                        stageCount={stageGroups.length}
+                        stageIdx={stageIdx}
+                        startIndex={startIndex}
+                      />
+                    );
+                  })}
+                  <NewStageDropZone
+                    hasEmptyStage={hasEmptyStage}
+                    onAddStage={() => setStageGroups((prev) => addEmptyStage(prev))}
+                  />
+                  {!stageGroups.length && (
+                    <p className="mat-stage-empty">
+                      从左侧点击或拖拽候选智能体加入编排，至少 {MIN_SEATS} 个。
+                    </p>
+                  )}
+                </div>
+              </div>
+            </div>
+
+            <DragOverlay>
+              {activeDrag ? (
+                <div className="mat-drag-overlay">
+                  <GripVertical size={14} />
+                  {activeDrag.name}
+                </div>
+              ) : null}
+            </DragOverlay>
+          </DndContext>
+        </section>
+      </div>
+
+      <Dialog
+        className="mat-settings-dialog"
+        closeBtn
+        closeOnOverlayClick={false}
+        destroyOnClose
+        footer={
+          <div className="mat-confirm-footer">
+            <Button variant="outline" onClick={() => setSettingsOpen(false)}>
+              取消
+            </Button>
+            <Button theme="primary" onClick={() => setSettingsOpen(false)}>
+              确定
+            </Button>
+          </div>
+        }
+        header="执行设置"
+        placement="top"
+        visible={settingsOpen}
+        width={620}
+        onClose={() => setSettingsOpen(false)}
+      >
+        <div className="mat-settings">
+          <div className="mat-field">
+            <span className="mat-field-label">执行频率</span>
+            <SchedulePicker value={schedule} onChange={setSchedule} />
+          </div>
+          <label className="mat-field">
+            <span className="mat-field-label">
+              任务指令<i>*</i>
+            </span>
+            <Textarea
+              autosize={{ minRows: 4, maxRows: 8 }}
+              maxlength={2000}
+              placeholder="描述每次定时执行要让智能体编队完成的事情，例如：汇总过去 24 小时重点舆情并生成早报…"
+              value={prompt}
+              onChange={(value) => setPrompt(String(value ?? ""))}
+            />
+          </label>
+          <div className="mat-field mat-field--switch">
+            <span className="mat-field-label">保存后立即启用</span>
+            <Switch value={enabled} onChange={(value) => setEnabled(!!value)} />
+          </div>
+        </div>
+      </Dialog>
+
+      <Dialog
+        className="mat-confirm-dialog"
+        closeBtn
+        closeOnOverlayClick={false}
+        destroyOnClose
+        footer={
+          <div className="mat-confirm-footer">
+            <Button variant="outline" onClick={() => setConfirmOpen(false)}>
+              再改改
+            </Button>
+            <Button loading={saving} theme="primary" onClick={() => void submitTask()}>
+              确认保存
+            </Button>
+          </div>
+        }
+        header="确认保存任务"
+        placement="top"
+        visible={confirmOpen}
+        width={720}
+        onClose={() => setConfirmOpen(false)}
+      >
+        <div className="mat-confirm">
+          <div className="mat-confirm-grid">
+            <div>
+              <span>任务名称</span>
+              <strong>{name.trim() || "—"}</strong>
+            </div>
+            <div>
+              <span>执行频率</span>
+              <strong>{describeCron(buildCron(schedule))}</strong>
+            </div>
+            <div>
+              <span>启用状态</span>
+              <strong>{enabled ? "启用" : "暂停"}</strong>
+            </div>
+            <div>
+              <span>智能体 / 阶段</span>
+              <strong>
+                {previewSeats.length} 个智能体 · {previewStages.length} 个阶段
+              </strong>
+            </div>
+          </div>
+          <div className="mat-confirm-flow">
+            <AgentFlow
+              seats={previewSeats}
+              showGoals
+              size="sm"
+              stageGoals={previewGroups.map((group) => (group.goal ?? "").trim() || null)}
+              stages={previewStages}
+            />
+          </div>
+          <p className="mat-confirm-legend">
+            <span>
+              <i className="mat-flow-mode is-parallel" style={{ cursor: "default" }} /> 并行：同阶段智能体同时执行
+            </span>
+            <span>
+              <i className="mat-flow-mode is-serial" style={{ cursor: "default" }} /> 串行：阶段之间依次执行
+            </span>
+          </p>
+          {prompt.trim() && (
+            <div className="mat-confirm-prompt">
+              <span>任务指令</span>
+              <p>{prompt.trim()}</p>
+            </div>
+          )}
+        </div>
+      </Dialog>
+
+      <AgentDetailDrawer
+        target={detailTarget}
+        onClose={() => setDetailTarget(null)}
+      />
+    </section>
+  );
+}

frontend-web/src/pages/official-multi-agent-tasks/components/AgentDetailDrawer.tsx

@@ -0,0 +1,164 @@
+/**
+ * 智能体详情抽屉：候选池卡片 / 编排席位条「查看详情」共用。
+ * 后端查不到（本地演示席位）时回退到编排里保存的名称/描述快照。
+ */
+
+import { Bot } from "lucide-react";
+import { Drawer, Tag } from "tdesign-react";
+
+import { useAgent } from "@/core/agents";
+
+export interface AgentDetailTarget {
+  agentId: string;
+  fallbackName?: string;
+  fallbackDescription?: string;
+}
+
+function formatDateTime(value: string | null | undefined): string {
+  if (!value) return "—";
+  const date = new Date(value);
+  if (Number.isNaN(date.getTime())) return "—";
+  return new Intl.DateTimeFormat("zh-CN", {
+    year: "numeric",
+    month: "2-digit",
+    day: "2-digit",
+    hour: "2-digit",
+    minute: "2-digit",
+    hour12: false,
+  }).format(date);
+}
+
+export default function AgentDetailDrawer({
+  target,
+  onClose,
+}: {
+  target: AgentDetailTarget | null;
+  onClose: () => void;
+}) {
+  const { agent, isLoading } = useAgent(target?.agentId ?? null, {
+    enabled: !!target,
+  });
+
+  const name = agent?.name ?? target?.fallbackName ?? target?.agentId ?? "";
+  const description =
+    agent?.description ?? target?.fallbackDescription ?? agent?.description ?? "";
+
+  return (
+    <Drawer
+      className="mat-agent-drawer"
+      closeBtn
+      destroyOnClose
+      footer={null}
+      header={
+        <div className="mat-agent-header">
+          <span className="mat-agent-glyph" aria-hidden="true">
+            <Bot size={17} />
+          </span>
+          <div className="mat-agent-heading">
+            <h3>{name}</h3>
+            <p>智能体详情</p>
+          </div>
+          {agent && (
+            <Tag size="small" theme={agent.builtin ? "default" : "primary"} variant="light">
+              {agent.builtin ? "内置" : agent.published ? "已发布" : "私有"}
+            </Tag>
+          )}
+        </div>
+      }
+      placement="right"
+      size="460px"
+      visible={!!target}
+      onClose={onClose}
+    >
+      {target && (
+        <div className="mat-agent">
+          {isLoading ? (
+            <p className="mat-agent-loading">正在加载智能体详情…</p>
+          ) : agent ? (
+            <>
+              <div className="mat-detail-grid">
+                <div>
+                  <span>创建者</span>
+                  <strong>
+                    {agent.owner_username
+                      ? agent.owner_username
+                      : agent.builtin
+                        ? "内置智能体"
+                        : "—"}
+                  </strong>
+                </div>
+                <div>
+                  <span>执行模型</span>
+                  <strong>{agent.model || "默认模型"}</strong>
+                </div>
+                <div>
+                  <span>创建时间</span>
+                  <strong>{formatDateTime(agent.created_at)}</strong>
+                </div>
+                <div>
+                  <span>技能数</span>
+                  <strong>{agent.skills?.length ?? 0} 个</strong>
+                </div>
+              </div>
+
+              {agent.description && (
+                <section className="mat-agent-section">
+                  <header className="mat-detail-section-head">
+                    <h4>简介</h4>
+                  </header>
+                  <p className="mat-agent-desc">{agent.description}</p>
+                </section>
+              )}
+
+              {!!agent.tags?.length && (
+                <section className="mat-agent-section">
+                  <header className="mat-detail-section-head">
+                    <h4>标签</h4>
+                  </header>
+                  <div className="mat-agent-tags">
+                    {agent.tags.map((tag) => (
+                      <Tag key={tag.id} size="small" variant="light">
+                        {tag.name}
+                      </Tag>
+                    ))}
+                  </div>
+                </section>
+              )}
+
+              {!!agent.skills?.length && (
+                <section className="mat-agent-section">
+                  <header className="mat-detail-section-head">
+                    <h4>已配置技能</h4>
+                  </header>
+                  <div className="mat-agent-tags">
+                    {agent.skills.map((skill) => (
+                      <Tag key={skill} size="small" theme="primary" variant="light">
+                        {skill}
+                      </Tag>
+                    ))}
+                  </div>
+                </section>
+              )}
+            </>
+          ) : (
+            <div className="mat-agent-local">
+              <div className="mat-detail-grid">
+                <div>
+                  <span>创建者</span>
+                  <strong>本地演示</strong>
+                </div>
+                <div>
+                  <span>数据来源</span>
+                  <strong>编排内置快照</strong>
+                </div>
+              </div>
+              <p className="mat-agent-desc">
+                {description || "该席位来自演示种子数据，未对应后端智能体目录中的记录。接入后端后，编排将引用真实智能体。"}
+              </p>
+            </div>
+          )}
+        </div>
+      )}
+    </Drawer>
+  );
+}

frontend-web/src/pages/official-multi-agent-tasks/components/AgentFlow.tsx

@@ -0,0 +1,136 @@
+import { Bot, MoveRight } from "lucide-react";
+import { Fragment } from "react";
+
+import type { MatSeat, MatSeatRunState } from "../types";
+import { seatColorIndex } from "../utils";
+
+export interface AgentFlowProps {
+  stages: string[][];
+  seats: MatSeat[];
+  stageGoals?: (string | null)[] | null;
+  /** 提供时在席位上渲染执行状态点（运行图例）。 */
+  seatStates?: Record<string, MatSeatRunState> | null;
+  /** 当前执行到的阶段下标；提供时高亮进度。 */
+  activeStageIndex?: number | null;
+  size?: "sm" | "md";
+  showGoals?: boolean;
+}
+
+const STATE_LABEL: Record<MatSeatRunState, string> = {
+  waiting: "等待",
+  running: "执行中",
+  done: "完成",
+  failed: "失败",
+};
+
+export default function AgentFlow({
+  stages,
+  seats,
+  stageGoals,
+  seatStates,
+  activeStageIndex,
+  size = "md",
+  showGoals = false,
+}: AgentFlowProps) {
+  const seatById = new Map(seats.map((seat) => [seat.agent_id, seat]));
+
+  // 只有一个阶段时不画阶段框：席位直接横向排列。
+  if (stages.length === 1) {
+    return (
+      <div className={`mat-flow mat-flow--single mat-flow--${size}`}>
+        {stages[0]!.map((agentId) => {
+          const seat = seatById.get(agentId);
+          const state = seatStates?.[agentId] ?? null;
+          return (
+            <div
+              className={`mat-seat mat-seat--c${seatColorIndex(agentId)}${
+                state ? ` is-${state}` : ""
+              }`}
+              key={agentId}
+              title={
+                seat?.description
+                  ? `${seat.name}：${seat.description}`
+                  : (seat?.name ?? agentId)
+              }
+            >
+              <span className="mat-seat-avatar">
+                <Bot size={size === "sm" ? 11 : 12} strokeWidth={2.2} />
+              </span>
+              <span className="mat-seat-name">{seat?.name ?? agentId}</span>
+              {state && (
+                <span className={`mat-seat-dot is-${state}`} title={STATE_LABEL[state]} />
+              )}
+            </div>
+          );
+        })}
+      </div>
+    );
+  }
+
+  return (
+    <div className={`mat-flow mat-flow--${size}`}>
+      {stages.map((stage, stageIndex) => {
+        const parallel = stage.length > 1;
+        const isActive = typeof activeStageIndex === "number" && activeStageIndex === stageIndex;
+        const isPast =
+          typeof activeStageIndex === "number" && activeStageIndex > stageIndex;
+        const goal = showGoals ? stageGoals?.[stageIndex] ?? null : null;
+        return (
+          <Fragment key={`stage-${stageIndex}`}>
+            {stageIndex > 0 && (
+              <div className="mat-flow-arrow" aria-hidden="true">
+                <MoveRight size={size === "sm" ? 14 : 16} strokeWidth={2} />
+              </div>
+            )}
+            <section
+              className={[
+                "mat-flow-stage",
+                isActive ? "is-active" : "",
+                isPast ? "is-done" : "",
+              ]
+                .filter(Boolean)
+                .join(" ")}
+            >
+              <header className="mat-flow-stage-head">
+                <span className="mat-flow-stage-no">阶段 {stageIndex + 1}</span>
+                {parallel ? (
+                  <span className="mat-flow-mode is-parallel">并行 ×{stage.length}</span>
+                ) : (
+                  <span className="mat-flow-mode is-serial">串行</span>
+                )}
+              </header>
+              <div className="mat-flow-seats">
+                {stage.map((agentId) => {
+                  const seat = seatById.get(agentId);
+                  const state = seatStates?.[agentId] ?? null;
+                  return (
+                    <div
+                      className={`mat-seat mat-seat--c${seatColorIndex(agentId)}${
+                        state ? ` is-${state}` : ""
+                      }`}
+                      key={agentId}
+                      title={
+                        seat?.description
+                          ? `${seat.name}：${seat.description}`
+                          : (seat?.name ?? agentId)
+                      }
+                    >
+                      <span className="mat-seat-avatar">
+                        <Bot size={size === "sm" ? 11 : 12} strokeWidth={2.2} />
+                      </span>
+                      <span className="mat-seat-name">{seat?.name ?? agentId}</span>
+                      {state && (
+                        <span className={`mat-seat-dot is-${state}`} title={STATE_LABEL[state]} />
+                      )}
+                    </div>
+                  );
+                })}
+              </div>
+              {goal && <p className="mat-flow-goal">{goal}</p>}
+            </section>
+          </Fragment>
+        );
+      })}
+    </div>
+  );
+}

frontend-web/src/pages/official-multi-agent-tasks/components/SchedulePicker.tsx

@@ -0,0 +1,76 @@
+import { Select } from "tdesign-react";
+
+import type { ScheduleConfig, ScheduleFreq } from "@/components/workspace/scheduled-tasks/schedule-builder";
+
+import { describeScheduleConfig } from "../utils";
+
+const WEEKDAYS = ["周日", "周一", "周二", "周三", "周四", "周五", "周六"];
+const HOURS = Array.from({ length: 24 }, (_, index) => index);
+const MINUTES = [0, 15, 30, 45];
+const DATES = Array.from({ length: 31 }, (_, index) => index + 1);
+
+const FREQ_OPTIONS: Array<{ label: string; value: ScheduleFreq }> = [
+  { label: "每小时", value: "hourly" },
+  { label: "每天", value: "daily" },
+  { label: "每周", value: "weekly" },
+  { label: "每月", value: "monthly" },
+];
+
+export default function SchedulePicker({
+  value,
+  onChange,
+}: {
+  value: ScheduleConfig;
+  onChange: (config: ScheduleConfig) => void;
+}) {
+  const set = (patch: Partial<ScheduleConfig>) => onChange({ ...value, ...patch });
+  const toNumber = (input: unknown) => Number(input as string | number);
+
+  return (
+    <div className="mat-schedule-picker">
+      <div className="mat-schedule-row">
+        <Select
+          options={FREQ_OPTIONS}
+          value={value.freq}
+          onChange={(next) => set({ freq: next as ScheduleFreq })}
+        />
+        {value.freq === "hourly" && (
+          <Select
+            options={MINUTES.map((minute) => ({ label: `${minute} 分`, value: minute }))}
+            value={value.minute}
+            onChange={(next) => set({ minute: toNumber(next) })}
+          />
+        )}
+        {value.freq === "weekly" && (
+          <Select
+            options={WEEKDAYS.map((label, index) => ({ label, value: index }))}
+            value={value.weekday}
+            onChange={(next) => set({ weekday: toNumber(next) })}
+          />
+        )}
+        {value.freq === "monthly" && (
+          <Select
+            options={DATES.map((date) => ({ label: `第 ${date} 日`, value: date }))}
+            value={value.date}
+            onChange={(next) => set({ date: toNumber(next) })}
+          />
+        )}
+        {value.freq !== "hourly" && (
+          <Select
+            options={HOURS.map((hour) => ({ label: `${hour} 时`, value: hour }))}
+            value={value.hour}
+            onChange={(next) => set({ hour: toNumber(next) })}
+          />
+        )}
+        {value.freq !== "hourly" && (
+          <Select
+            options={MINUTES.map((minute) => ({ label: `${minute} 分`, value: minute }))}
+            value={value.minute}
+            onChange={(next) => set({ minute: toNumber(next) })}
+          />
+        )}
+      </div>
+      <p className="mat-schedule-preview">{describeScheduleConfig(value)} 自动执行</p>
+    </div>
+  );
+}

frontend-web/src/pages/official-multi-agent-tasks/store.ts

@@ -0,0 +1,361 @@
+/**
+ * 多智能体定时任务的本地 mock 存储层。
+ *
+ * 后端接口尚未实现，先用 localStorage 持久化任务与运行记录：
+ * - 「立即执行」写入一条运行记录，状态由 started_at + duration_ms 在
+ *   读取时按当前时间推导（排队 → 按阶段推进 → 成功/失败），配合
+ *   React Query 的轮询即可获得无后端的“实时”执行动画。
+ * - 接口就绪后，把本文件的 CRUD 换成 HTTP 调用即可，页面无需改动。
+ */
+
+import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query";
+
+import { uuid } from "@/core/utils/uuid";
+
+import { expandStages, nextCronDate } from "./utils";
+import type {
+  MatRunRecord,
+  MatRunStatus,
+  MatRunView,
+  MatSeatRunState,
+  MatTask,
+  MatTaskPayload,
+  MatTaskView,
+} from "./types";
+
+const STORAGE_KEY = "deerflow.official.multi-agent-tasks.v1";
+
+interface MatStore {
+  tasks: MatTask[];
+  runs: MatRunRecord[];
+}
+
+const PENDING_MS = 1_500;
+const STAGE_MS = 4_000;
+
+function readStore(): MatStore {
+  try {
+    const raw = window.localStorage.getItem(STORAGE_KEY);
+    if (raw) {
+      const parsed = JSON.parse(raw) as Partial<MatStore>;
+      if (Array.isArray(parsed.tasks)) {
+        return { tasks: parsed.tasks as MatTask[], runs: parsed.runs ?? [] };
+      }
+    }
+  } catch {
+    // 解析失败按首次访问处理，重写种子数据。
+  }
+  const seeded = seedStore();
+  writeStore(seeded);
+  return seeded;
+}
+
+function writeStore(store: MatStore) {
+  try {
+    window.localStorage.setItem(STORAGE_KEY, JSON.stringify(store));
+  } catch {
+    // 存储被禁用时静默降级为内存态。
+  }
+}
+
+function mutateStore(mutator: (store: MatStore) => MatStore): MatStore {
+  const next = mutator(readStore());
+  writeStore(next);
+  return next;
+}
+
+function hoursAgo(hours: number): string {
+  return new Date(Date.now() - hours * 3_600_000).toISOString();
+}
+
+function seedStore(): MatStore {
+  const seats = {
+    collector: { agent_id: "mat-seed-collector", name: "全网舆情采集", description: "多源检索并汇总最新舆情线索" },
+    monitor: { agent_id: "mat-seed-monitor", name: "舆情监测雷达", description: "盯住重点信源并输出监测明细" },
+    analyst: { agent_id: "mat-seed-analyst", name: "舆情研判分析", description: "对采集结果做情感与趋势研判" },
+    writer: { agent_id: "mat-seed-writer", name: "报告撰写助手", description: "把各阶段产出整理成结构化报告" },
+    verifier: { agent_id: "mat-seed-verifier", name: "数据核查员", description: "交叉校验数据与结论的可靠性" },
+    rival: { agent_id: "mat-seed-rival", name: "竞品情报员", description: "跟踪竞品动态与公开信息" },
+    researcher: { agent_id: "mat-seed-researcher", name: "行业研究员", description: "梳理行业政策与市场变化" },
+    forecaster: { agent_id: "mat-seed-forecaster", name: "趋势预测专家", description: "基于数据推演后续走势" },
+    reviewer: { agent_id: "mat-seed-reviewer", name: "内容审核官", description: "审核报告口径与合规风险" },
+  } satisfies Record<string, MatTask["seats"][number]>;
+
+  const tasks: MatTask[] = [
+    {
+      id: "mat-task-daily",
+      name: "每日舆情早报",
+      description: "每天早上自动汇总过去 24 小时重点舆情，并行采集 + 研判 + 成稿。",
+      prompt:
+        "汇总过去 24 小时与本公司相关的重点舆情：先并行完成多源采集与监测，再研判情感与趋势，最后汇总为一份不超过 1500 字的舆情早报，附风险提示。",
+      coordinator_prompt: "采集阶段允许两个智能体各自独立检索，不做信息合并；研判阶段需引用采集阶段的原始线索。",
+      cron_expr: "0 9 * * *",
+      timezone: "Asia/Shanghai",
+      enabled: true,
+      seats: [seats.collector, seats.monitor, seats.analyst, seats.writer],
+      stages: [
+        [seats.collector.agent_id, seats.monitor.agent_id],
+        [seats.analyst.agent_id],
+        [seats.writer.agent_id],
+      ],
+      stage_goals: ["并行完成全网采集与重点信源监测", "基于原始线索完成情感与趋势研判", "汇总为可直发的舆情早报"],
+      created_at: hoursAgo(24 * 18),
+      updated_at: hoursAgo(26),
+    },
+    {
+      id: "mat-task-weekly",
+      name: "周度竞品情报速览",
+      description: "每周一串行走查竞品动态，核查后输出速览简报。",
+      prompt: "跟踪 5 家核心竞品过去一周的公开动态（产品、市场、人事），核查数据可信度后输出竞品速览简报。",
+      cron_expr: "30 8 * * 1",
+      timezone: "Asia/Shanghai",
+      enabled: false,
+      seats: [seats.rival, seats.verifier, seats.writer],
+      stages: null,
+      stage_goals: null,
+      created_at: hoursAgo(24 * 40),
+      updated_at: hoursAgo(24 * 3),
+    },
+    {
+      id: "mat-task-monthly",
+      name: "月度行业洞察报告",
+      description: "每月并行完成行业研究、趋势预测与数据核查，审核后成稿。",
+      prompt:
+        "围绕所在行业输出月度洞察：行业研究员、趋势预测专家与数据核查员并行推进，随后由内容审核官把关，最终产出月度洞察报告。",
+      cron_expr: "0 10 1 * *",
+      timezone: "Asia/Shanghai",
+      enabled: true,
+      seats: [seats.researcher, seats.forecaster, seats.verifier, seats.reviewer, seats.writer],
+      stages: [
+        [seats.researcher.agent_id, seats.forecaster.agent_id, seats.verifier.agent_id],
+        [seats.reviewer.agent_id],
+        [seats.writer.agent_id],
+      ],
+      stage_goals: ["并行完成行业研究、趋势预测与数据核查", "审核结论口径与合规风险", "整合为月度洞察报告"],
+      created_at: hoursAgo(24 * 9),
+      updated_at: hoursAgo(24 * 9),
+    },
+  ];
+
+  const runs: MatRunRecord[] = [
+    runSeed("mat-task-daily", 2.2, 3, null),
+    runSeed("mat-task-daily", 26, 3, null),
+    runSeed("mat-task-daily", 50, 3, null),
+    runSeed("mat-task-weekly", 24 * 3, 3, 2),
+    runSeed("mat-task-weekly", 24 * 10, 3, null),
+  ];
+  return { tasks, runs };
+}
+
+function runSeed(
+  taskId: string,
+  hoursAgoValue: number,
+  stageTotal: number,
+  failStage: number | null,
+): MatRunRecord {
+  const started = hoursAgo(hoursAgoValue);
+  return {
+    id: uuid() as string,
+    task_id: taskId,
+    trigger: "schedule",
+    scheduled_for: started,
+    started_at: started,
+    duration_ms: PENDING_MS + stageTotal * STAGE_MS,
+    fail_stage: failStage,
+  };
+}
+
+/** 由记录 + 当前时间推导运行视图：排队 → 按阶段推进 → 成功/失败。 */
+function resolveRunView(run: MatRunRecord, stages: string[][], now = Date.now()): MatRunView {
+  const startedAt = new Date(run.started_at).getTime();
+  const elapsed = now - startedAt;
+  const stageTotal = Math.max(1, stages.length);
+  const duration = run.duration_ms || PENDING_MS + stageTotal * STAGE_MS;
+
+  const buildStates = (
+    status: MatRunStatus,
+    activeStage: number,
+  ): Record<string, MatSeatRunState> => {
+    const states: Record<string, MatSeatRunState> = {};
+    stages.forEach((stage, index) => {
+      const seatState: MatSeatRunState =
+        status === "succeeded"
+          ? "done"
+          : index < activeStage
+            ? "done"
+            : index === activeStage
+              ? status === "failed"
+                ? "failed"
+                : "running"
+              : "waiting";
+      stage.forEach((agentId) => {
+        states[agentId] = seatState;
+      });
+    });
+    return states;
+  };
+
+  if (elapsed < PENDING_MS) {
+    return {
+      ...run,
+      finished_at: null,
+      status: "pending",
+      stage_index: 0,
+      stage_total: stageTotal,
+      seat_states: buildStates("pending", -1),
+      error: null,
+    };
+  }
+
+  const failStage = typeof run.fail_stage === "number" ? run.fail_stage : null;
+  const progress = Math.min(elapsed - PENDING_MS, duration - PENDING_MS);
+  const activeStage = Math.min(Math.floor(progress / STAGE_MS), stageTotal - 1);
+
+  if (failStage !== null && activeStage >= failStage) {
+    const finishedAt = new Date(startedAt + PENDING_MS + (failStage + 1) * STAGE_MS).toISOString();
+    return {
+      ...run,
+      finished_at: finishedAt,
+      status: "failed",
+      stage_index: failStage,
+      stage_total: stageTotal,
+      seat_states: buildStates("failed", failStage),
+      error: `阶段 ${failStage + 1} 执行失败：模拟的执行异常（前端演示数据）`,
+    };
+  }
+
+  if (elapsed >= duration) {
+    return {
+      ...run,
+      finished_at: new Date(startedAt + duration).toISOString(),
+      status: "succeeded",
+      stage_index: stageTotal - 1,
+      stage_total: stageTotal,
+      seat_states: buildStates("succeeded", stageTotal),
+      error: null,
+    };
+  }
+
+  return {
+    ...run,
+    finished_at: null,
+    status: "running",
+    stage_index: activeStage,
+    stage_total: stageTotal,
+    seat_states: buildStates("running", activeStage),
+    error: null,
+  };
+}
+
+function buildTaskViews(now = Date.now()): MatTaskView[] {
+  const store = readStore();
+  return store.tasks.map((task) => {
+    const stages = expandStages(task);
+    const runs = store.runs
+      .filter((run) => run.task_id === task.id)
+      .map((run) => resolveRunView(run, stages, now))
+      .sort((a, b) => (a.started_at < b.started_at ? 1 : -1));
+    const next = task.enabled ? nextCronDate(task.cron_expr, new Date(now)) : null;
+    return {
+      ...task,
+      runs,
+      lastRun: runs[0] ?? null,
+      nextRunAt: next ? next.toISOString() : null,
+    };
+  });
+}
+
+function anyRunLive(tasks: MatTaskView[] | undefined): boolean {
+  return !!tasks?.some((task) =>
+    task.runs.some((run) => run.status === "pending" || run.status === "running"),
+  );
+}
+
+export function useMultiAgentTasks() {
+  const query = useQuery({
+    queryKey: ["mat", "tasks"],
+    queryFn: () => buildTaskViews(),
+    refetchInterval: (q) => (anyRunLive(q.state.data as MatTaskView[] | undefined) ? 1_800 : false),
+  });
+  return { tasks: query.data ?? [], isLoading: query.isLoading, error: query.error };
+}
+
+export function useMultiAgentTaskActions() {
+  const queryClient = useQueryClient();
+  const invalidate = () => queryClient.invalidateQueries({ queryKey: ["mat", "tasks"] });
+
+  const create = useMutation({
+    mutationFn: async (payload: MatTaskPayload) => {
+      const now = new Date().toISOString();
+      const task: MatTask = { ...payload, id: uuid() as string, created_at: now, updated_at: now };
+      mutateStore((store) => ({ ...store, tasks: [task, ...store.tasks] }));
+      return task;
+    },
+    onSuccess: invalidate,
+  });
+
+  const update = useMutation({
+    mutationFn: async ({ id, payload }: { id: string; payload: MatTaskPayload }) => {
+      mutateStore((store) => ({
+        ...store,
+        tasks: store.tasks.map((task) =>
+          task.id === id ? { ...task, ...payload, updated_at: new Date().toISOString() } : task,
+        ),
+      }));
+    },
+    onSuccess: invalidate,
+  });
+
+  const remove = useMutation({
+    mutationFn: async (id: string) => {
+      mutateStore((store) => ({
+        tasks: store.tasks.filter((task) => task.id !== id),
+        runs: store.runs.filter((run) => run.task_id !== id),
+      }));
+    },
+    onSuccess: invalidate,
+  });
+
+  const setEnabled = useMutation({
+    mutationFn: async ({ id, enabled }: { id: string; enabled: boolean }) => {
+      mutateStore((store) => ({
+        ...store,
+        tasks: store.tasks.map((task) =>
+          task.id === id ? { ...task, enabled, updated_at: new Date().toISOString() } : task,
+        ),
+      }));
+    },
+    onSuccess: invalidate,
+  });
+
+  const trigger = useMutation({
+    mutationFn: async (id: string) => {
+      const store = readStore();
+      const task = store.tasks.find((item) => item.id === id);
+      if (!task) throw new Error("任务不存在");
+      const stageTotal = expandStages(task).length;
+      const now = new Date().toISOString();
+      const run: MatRunRecord = {
+        id: uuid() as string,
+        task_id: id,
+        trigger: "manual",
+        scheduled_for: now,
+        started_at: now,
+        duration_ms: PENDING_MS + stageTotal * STAGE_MS,
+        fail_stage: null,
+      };
+      mutateStore((current) => ({ ...current, runs: [run, ...current.runs] }));
+      return run;
+    },
+    onSuccess: invalidate,
+  });
+
+  const removeRun = useMutation({
+    mutationFn: async (runId: string) => {
+      mutateStore((store) => ({ ...store, runs: store.runs.filter((run) => run.id !== runId) }));
+    },
+    onSuccess: invalidate,
+  });
+
+  return { create, update, remove, setEnabled, trigger, removeRun };
+}

frontend-web/src/pages/official-multi-agent-tasks/types.ts

@@ -0,0 +1,92 @@
+/**
+ * 多智能体定时任务 —— 前端先行数据模型。
+ *
+ * 编排语义与业务链条保持一致：同一阶段（stage）内的多个智能体并行执行，
+ * 阶段之间串行执行；`stages: null` 表示纯串行线性链（每个智能体一层）。
+ * 当前后端尚未提供接口，数据先落 localStorage，接口就绪后只需替换 store 层。
+ */
+
+export type MatRunStatus = "pending" | "running" | "succeeded" | "failed";
+export type MatSeatRunState = "waiting" | "running" | "done" | "failed";
+export type MatRunTrigger = "manual" | "schedule";
+
+/** 编排中的一个智能体席位（落库时保存名称快照，展示优先用快照）。 */
+export interface MatSeat {
+  agent_id: string;
+  name: string;
+  description?: string;
+}
+
+export interface MatTask {
+  id: string;
+  name: string;
+  description?: string;
+  /** 任务指令：本次定时执行要让智能体编队完成什么。 */
+  prompt: string;
+  /** 总控（协调者）提示词：告诉总控如何编排各阶段的产出。 */
+  coordinator_prompt?: string;
+  cron_expr: string;
+  timezone: string;
+  enabled: boolean;
+  seats: MatSeat[];
+  /** `stages[i]` 内的智能体并行，i 递增串行；null = 线性串行。 */
+  stages: string[][] | null;
+  /** 与 stages 同下标对齐的每阶段目标。 */
+  stage_goals?: (string | null)[] | null;
+  created_at: string;
+  updated_at: string;
+}
+
+/** 落库的运行记录；状态由 started_at + duration_ms 在读取时推导。 */
+export interface MatRunRecord {
+  id: string;
+  task_id: string;
+  trigger: MatRunTrigger;
+  scheduled_for: string;
+  started_at: string;
+  duration_ms: number;
+  /** 模拟在该阶段（下标）失败；null = 成功。 */
+  fail_stage?: number | null;
+}
+
+/** 读取时推导出的运行视图。 */
+export interface MatRunView {
+  id: string;
+  task_id: string;
+  trigger: MatRunTrigger;
+  scheduled_for: string;
+  started_at: string;
+  finished_at: string | null;
+  status: MatRunStatus;
+  /** 当前执行到的阶段下标（从 0 开始）。 */
+  stage_index: number;
+  stage_total: number;
+  seat_states: Record<string, MatSeatRunState>;
+  error: string | null;
+}
+
+export interface MatTaskView extends MatTask {
+  runs: MatRunView[];
+  lastRun: MatRunView | null;
+  nextRunAt: string | null;
+}
+
+/** 编辑器内部使用的阶段草稿。 */
+export interface MatStageGroup {
+  seats: MatSeat[];
+  goal: string;
+}
+
+/** 编辑器保存时对外提交的载荷。 */
+export interface MatTaskPayload {
+  name: string;
+  description?: string;
+  prompt: string;
+  coordinator_prompt?: string;
+  cron_expr: string;
+  timezone: string;
+  enabled: boolean;
+  seats: MatSeat[];
+  stages: string[][] | null;
+  stage_goals: (string | null)[] | null;
+}

frontend-web/src/pages/official-multi-agent-tasks/utils.ts

@@ -0,0 +1,166 @@
+import {
+  DEFAULT_SCHEDULE_CONFIG,
+  type ScheduleConfig,
+  parseCron,
+} from "@/components/workspace/scheduled-tasks/schedule-builder";
+
+import type { MatSeat, MatTask, MatTaskView } from "./types";
+
+const WEEKDAY_LABELS = ["周日", "周一", "周二", "周三", "周四", "周五", "周六"];
+
+function pad2(value: number) {
+  return String(value).padStart(2, "0");
+}
+
+/** 把任务里的 stages 语义展开成显式阶段列表；null = 每个席位独立一层。 */
+export function expandStages(task: Pick<MatTask, "seats" | "stages">): string[][] {
+  const known = new Set(task.seats.map((seat) => seat.agent_id));
+  if (!task.stages?.length) {
+    return task.seats.map((seat) => [seat.agent_id]);
+  }
+  const expanded = task.stages
+    .map((stage) => stage.filter((id) => known.has(id)))
+    .filter((stage) => stage.length > 0);
+  return expanded.length ? expanded : task.seats.map((seat) => [seat.agent_id]);
+}
+
+export function describeScheduleConfig(cfg: ScheduleConfig): string {
+  const clock = `${pad2(cfg.hour)}:${pad2(cfg.minute)}`;
+  switch (cfg.freq) {
+    case "hourly":
+      return `每小时 ${cfg.minute} 分`;
+    case "weekly":
+      return `每${WEEKDAY_LABELS[cfg.weekday] ?? "周一"} ${clock}`;
+    case "monthly":
+      return `每月 ${cfg.date} 日 ${clock}`;
+    default:
+      return `每天 ${clock}`;
+  }
+}
+
+/** 由 cron 反推人类可读的执行频率描述（覆盖本页生成的所有 cron 形态）。 */
+export function describeCron(cron: string | null | undefined): string {
+  const cfg = parseCron(cron ?? undefined);
+  if (!cfg) return cron || "未设置";
+  return describeScheduleConfig(cfg);
+}
+
+/** 计算下一次执行时间；仅支持本页 schedule picker 生成的 5 段 cron 形态。 */
+export function nextCronDate(cron: string | null | undefined, from = new Date()): Date | null {
+  const cfg = parseCron(cron ?? undefined);
+  if (!cfg) return null;
+  const candidate = new Date(from);
+  candidate.setSeconds(0, 0);
+  switch (cfg.freq) {
+    case "hourly": {
+      candidate.setMinutes(cfg.minute);
+      if (candidate <= from) candidate.setHours(candidate.getHours() + 1);
+      return candidate;
+    }
+    case "daily": {
+      candidate.setHours(cfg.hour, cfg.minute, 0, 0);
+      if (candidate <= from) candidate.setDate(candidate.getDate() + 1);
+      return candidate;
+    }
+    case "weekly": {
+      candidate.setHours(cfg.hour, cfg.minute, 0, 0);
+      const day = candidate.getDay();
+      let delta = (cfg.weekday - day + 7) % 7;
+      if (delta === 0 && candidate <= from) delta = 7;
+      candidate.setDate(candidate.getDate() + delta);
+      return candidate;
+    }
+    case "monthly": {
+      candidate.setHours(cfg.hour, cfg.minute, 0, 0);
+      candidate.setDate(cfg.date);
+      if (candidate <= from) candidate.setMonth(candidate.getMonth() + 1);
+      return candidate;
+    }
+    default:
+      return null;
+  }
+}
+
+export function formatDateTime(value: string | null | undefined): string {
+  if (!value) return "—";
+  const date = new Date(value);
+  if (Number.isNaN(date.getTime())) return "—";
+  return new Intl.DateTimeFormat("zh-CN", {
+    month: "2-digit",
+    day: "2-digit",
+    hour: "2-digit",
+    minute: "2-digit",
+    hour12: false,
+  }).format(date);
+}
+
+export function formatRelative(value: string | null | undefined): string {
+  if (!value) return "—";
+  const date = new Date(value);
+  if (Number.isNaN(date.getTime())) return "—";
+  const elapsed = Date.now() - date.getTime();
+  if (elapsed < 60_000) return "刚刚";
+  if (elapsed < 3_600_000) return `${Math.floor(elapsed / 60_000)} 分钟前`;
+  if (elapsed < 86_400_000) return `${Math.floor(elapsed / 3_600_000)} 小时前`;
+  if (elapsed < 7 * 86_400_000) return `${Math.floor(elapsed / 86_400_000)} 天前`;
+  return formatDateTime(value);
+}
+
+export function formatDuration(ms: number): string {
+  if (!Number.isFinite(ms) || ms <= 0) return "—";
+  const seconds = Math.round(ms / 1000);
+  if (seconds < 60) return `${seconds} 秒`;
+  const minutes = Math.floor(seconds / 60);
+  const rest = seconds % 60;
+  return rest ? `${minutes} 分 ${rest} 秒` : `${minutes} 分钟`;
+}
+
+export function runStatusLabel(status: string | null | undefined): string {
+  switch (status) {
+    case "succeeded":
+      return "执行成功";
+    case "failed":
+      return "执行失败";
+    case "running":
+      return "执行中";
+    case "pending":
+      return "排队中";
+    default:
+      return "暂无执行";
+  }
+}
+
+export type MatTaskBadge = {
+  label: string;
+  theme: "success" | "warning" | "danger" | "default";
+};
+
+/** 任务卡右上角的状态徽标：以最近一次执行状态优先，其次启停。 */
+export function taskBadgeOf(task: MatTaskView): MatTaskBadge {
+  const status = task.lastRun?.status;
+  if (status === "running" || status === "pending") {
+    return { label: "执行中", theme: "warning" };
+  }
+  if (!task.enabled) return { label: "已暂停", theme: "default" };
+  if (status === "failed") return { label: "上次失败", theme: "danger" };
+  if (status === "succeeded") return { label: "运行正常", theme: "success" };
+  return { label: "已启用", theme: "success" };
+}
+
+/** 席位配色槽位：按 agent_id 稳定散列到 0..7。 */
+export function seatColorIndex(agentId: string): number {
+  let hash = 0;
+  for (let i = 0; i < agentId.length; i += 1) {
+    hash = (hash * 31 + agentId.charCodeAt(i)) >>> 0;
+  }
+  return hash % 8;
+}
+
+export function seatIndexOf(seats: MatSeat[], agentId: string): number {
+  const index = seats.findIndex((seat) => seat.agent_id === agentId);
+  return index >= 0 ? index : seats.length;
+}
+
+export function defaultScheduleConfig(): ScheduleConfig {
+  return { ...DEFAULT_SCHEDULE_CONFIG };
+}

frontend-web/src/pages/official-task-analysis/TaskAnalysisDetailPage.tsx

@@ -0,0 +1,340 @@
+/**
+ * 任务分析 · 详情页：左侧执行记录（定时 / 手动），右侧按选中运行回放流程图
+ * （业务链风格节点状态 + 耗时，5 秒自动刷新），点击智能体节点打开执行过程抽屉。
+ */
+
+import {
+  ArrowLeft,
+  CalendarClock,
+  CircleAlert,
+  GitBranch,
+  MessageSquareText,
+  Pencil,
+  Pause,
+  Play,
+  RefreshCw,
+  Zap,
+} from "lucide-react";
+import { useEffect, useMemo, useState } from "react";
+import { useNavigate } from "react-router-dom";
+import { Button, Switch, Tag } from "tdesign-react";
+import { toast } from "sonner";
+
+import { OFFICIAL_DEERFLOW_ROUTE_PREFIX } from "@/core/official-deerflow/routes";
+import {
+  formatDateTime,
+  formatDuration,
+  formatRelative,
+} from "@/pages/official-multi-agent-tasks/utils";
+
+import BindingDialog from "./components/BindingDialog";
+import SeatProcessDrawer from "./components/SeatProcessDrawer";
+import { ChainFlowDiagram } from "@/roundtable-planning/components/ChainFlowDiagram";
+import { taskBadgeOf } from "./badge";
+import { useTaTaskActions, useTaTasks, useTaTaskRuns } from "./store";
+
+const RUN_STATUS_TEXT: Record<string, string> = {
+  succeeded: "成功",
+  failed: "失败",
+  running: "执行中",
+  pending: "排队中",
+};
+
+const RUN_STATUS_THEME: Record<string, "success" | "danger" | "warning" | "default"> = {
+  succeeded: "success",
+  failed: "danger",
+  running: "warning",
+  pending: "default",
+};
+
+export default function TaskAnalysisDetailPage({
+  routeTaskId,
+}: {
+  routeTaskId: string | null;
+}) {
+  const navigate = useNavigate();
+  const { views, isLoading } = useTaTasks();
+  const actions = useTaTaskActions();
+  const taskView = views.find(({ task }) => task.id === routeTaskId) ?? null;
+  const task = taskView?.task ?? null;
+  const { runs, isLoading: runsLoading } = useTaTaskRuns(task?.id ?? null);
+
+  const [selectedRunId, setSelectedRunId] = useState<string | null>(null);
+  const [seatAgentId, setSeatAgentId] = useState<string | null>(null);
+  const [bindingVisible, setBindingVisible] = useState(false);
+  const [busy, setBusy] = useState(false);
+
+  // 默认选中最近一次运行；新触发的运行自动跟随。
+  useEffect(() => {
+    if (!runs.length) return;
+    if (!selectedRunId || !runs.some((run) => run.id === selectedRunId)) {
+      setSelectedRunId(runs[0]!.id);
+    }
+  }, [runs, selectedRunId]);
+
+  const selectedRun = useMemo(
+    () => runs.find((run) => run.id === selectedRunId) ?? runs[0] ?? null,
+    [runs, selectedRunId],
+  );
+
+  const backToList = () => {
+    navigate(`${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/task-analysis`);
+  };
+
+  const runTrigger = async (manual: boolean) => {
+    if (!task) return;
+    setBusy(true);
+    try {
+      if (manual) {
+        await actions.triggerManualRun.mutateAsync(task.id);
+        toast.success("已发起手动执行（总控问答）");
+      } else {
+        await actions.triggerRun.mutateAsync(task.id);
+        toast.success("已触发定时执行");
+      }
+      setSelectedRunId(null);
+    } catch {
+      toast.error("任务未绑定分析方案");
+    } finally {
+      setBusy(false);
+    }
+  };
+
+  const runToggle = async () => {
+    if (!task?.binding) return;
+    setBusy(true);
+    try {
+      await actions.setEnabled.mutateAsync({
+        taskId: task.id,
+        enabled: !task.binding?.enabled,
+      });
+      toast.success(task.binding?.enabled ? "已暂停" : "已恢复");
+    } catch {
+      toast.error("操作失败，请重试");
+    } finally {
+      setBusy(false);
+    }
+  };
+
+  if (!isLoading && !task) {
+    return (
+      <section className="ta-page">
+        <div className="ta-state ta-state--empty">
+          <h3>任务不存在或已被删除</h3>
+          <Button theme="primary" onClick={backToList}>
+            返回任务列表
+          </Button>
+        </div>
+      </section>
+    );
+  }
+
+  const binding = task?.binding ?? null;
+  const taskBadge = task && taskView ? taskBadgeOf(task, taskView.runs) : null;
+  const live = selectedRun?.status === "running" || selectedRun?.status === "pending";
+
+  return (
+    <section className="ta-page ta-detail-page">
+      <header className="ta-detail-head">
+        <button
+          aria-label="返回任务列表"
+          className="ta-back"
+          onClick={backToList}
+          type="button"
+        >
+          <ArrowLeft size={18} />
+        </button>
+        <div className="ta-detail-heading">
+          <h1>
+            {task?.name ?? "…"}
+            <em>{task?.sourceTaskId}</em>
+          </h1>
+          <p>
+            {binding ? (
+              <>
+                <GitBranch aria-hidden="true" size={13} />
+                {binding.chainTitle} · {binding.cronText}
+                {binding.directionNotes.length > 0 &&
+                  ` · ${binding.directionNotes.length} 条方向备注`}
+              </>
+            ) : (
+              task?.description
+            )}
+          </p>
+        </div>
+        {taskBadge && (
+          <Tag theme={taskBadge.theme} variant="light">
+            {taskBadge.label}
+          </Tag>
+        )}
+        {binding && (
+          <Switch value={binding.enabled} onChange={() => void runToggle()} />
+        )}
+        <div className="ta-detail-ops">
+          {binding && (
+            <>
+              <Button
+                disabled={busy}
+                loading={busy}
+                theme="primary"
+                onClick={() => void runTrigger(false)}
+              >
+                <Zap aria-hidden="true" size={15} />
+                立即执行
+              </Button>
+              <Button variant="outline" disabled={busy} onClick={() => void runTrigger(true)}>
+                <MessageSquareText aria-hidden="true" size={15} />
+                模拟手动执行
+              </Button>
+            </>
+          )}
+          <Button variant="outline" onClick={() => setBindingVisible(true)}>
+            <Pencil aria-hidden="true" size={15} />
+            {binding ? "编辑绑定" : "配置分析"}
+          </Button>
+        </div>
+      </header>
+
+      {!binding ? (
+        <div className="ta-detail-empty">
+          <GitBranch aria-hidden="true" size={26} />
+          <h3>尚未配置分析方案</h3>
+          <p>绑定业务链与内容块映射后，这里将展示执行记录与各智能体的执行流程。</p>
+          <Button theme="primary" onClick={() => setBindingVisible(true)}>
+            配置分析
+          </Button>
+        </div>
+      ) : (
+        <div className="ta-detail-layout">
+          <aside className="ta-runs">
+            <header className="ta-runs-head">
+              <h2>执行记录</h2>
+              <span>
+                <RefreshCw aria-hidden="true" className={live ? "ta-spin" : undefined} size={12} />
+                每 5 秒刷新
+              </span>
+            </header>
+            <div className="ta-runs-list">
+              {runsLoading && !runs.length ? (
+                <p className="ta-runs-empty">正在加载执行记录…</p>
+              ) : runs.length ? (
+                runs.map((run) => {
+                  const active = selectedRun?.id === run.id;
+                  const duration = run.finishedAt
+                    ? formatDuration(
+                        new Date(run.finishedAt).getTime() - new Date(run.startedAt).getTime(),
+                      )
+                    : null;
+                  return (
+                    <button
+                      className={`ta-run-item${active ? " is-active" : ""}`}
+                      key={run.id}
+                      onClick={() => setSelectedRunId(run.id)}
+                      type="button"
+                    >
+                      <span className="ta-run-item-top">
+                        {run.trigger === "manual" ? (
+                          <MessageSquareText aria-hidden="true" size={13} />
+                        ) : (
+                          <CalendarClock aria-hidden="true" size={13} />
+                        )}
+                        <b>{run.trigger === "manual" ? "手动执行" : "定时执行"}</b>
+                        <Tag size="small" theme={RUN_STATUS_THEME[run.status] ?? "default"} variant="light">
+                          {RUN_STATUS_TEXT[run.status] ?? run.status}
+                        </Tag>
+                      </span>
+                      <span className="ta-run-item-time">
+                        {formatDateTime(run.startedAt)} · {duration ?? "进行中"}
+                      </span>
+                      {run.trigger === "manual" && run.manualSummary && (
+                        <span className="ta-run-item-summary">{run.manualSummary}</span>
+                      )}
+                      {run.trigger === "schedule" && (
+                        <span className="ta-run-item-blocks">
+                          内容块 {run.blockStates.filter((b) => b.state === "done").length}/
+                          {run.blockStates.length}
+                        </span>
+                      )}
+                    </button>
+                  );
+                })
+              ) : (
+                <p className="ta-runs-empty">还没有执行记录，点「立即执行」体验完整流程。</p>
+              )}
+            </div>
+          </aside>
+
+          <section className="ta-flow-panel">
+            {selectedRun ? (
+              <>
+                <header className="ta-flow-head">
+                  <h2>
+                    执行流程
+                    <small>
+                      {selectedRun.trigger === "manual" ? "手动执行" : "定时执行"} ·{" "}
+                      {formatRelative(selectedRun.startedAt)}
+                      {selectedRun.status === "running" &&
+                        ` · 阶段 ${Math.min(selectedRun.stageIndex + 1, selectedRun.stageTotal)}/${selectedRun.stageTotal} 执行中`}
+                    </small>
+                  </h2>
+                  <span className="ta-flow-legend">
+                    <i className="ta-dot is-done" /> 完成
+                    <i className="ta-dot is-running" /> 执行中
+                    <i className="ta-dot is-waiting" /> 等待
+                    <i className="ta-dot is-failed" /> 失败
+                  </span>
+                </header>
+                <div className="ta-flow-canvas">
+                  <ChainFlowDiagram
+                    className="ta-flow-diagram"
+                    onNodeClick={(agentId) => setSeatAgentId(agentId)}
+                    seatStates={selectedRun.seatInfos}
+                    stages={selectedRun.stages.map((stage, index) => ({
+                      seats: stage.map((agentId) => ({
+                        agent_id: agentId,
+                        name: selectedRun.seatNames[agentId] ?? agentId,
+                      })),
+                      goal: selectedRun.stageGoals?.[index] ?? null,
+                    }))}
+                  />
+                </div>
+                {selectedRun.error && (
+                  <p className="ta-flow-error">
+                    <CircleAlert aria-hidden="true" size={14} />
+                    {selectedRun.error}
+                  </p>
+                )}
+              </>
+            ) : (
+              <div className="ta-flow-panel-empty">选择左侧一条执行记录查看流程</div>
+            )}
+          </section>
+        </div>
+      )}
+
+      <SeatProcessDrawer
+        onClose={() => setSeatAgentId(null)}
+        run={selectedRun}
+        seatAgentId={seatAgentId}
+        task={task ?? undefined}
+      />
+
+      <BindingDialog
+        onClose={() => setBindingVisible(false)}
+        onSubmit={async (payload) => {
+          if (!task) return;
+          try {
+            await actions.bind.mutateAsync({ taskId: task.id, payload });
+            toast.success(binding ? "绑定已更新" : "分析方案已绑定");
+            setBindingVisible(false);
+          } catch {
+            toast.error("保存失败，请重试");
+          }
+        }}
+        submitting={actions.bind.isPending}
+        task={task}
+        visible={bindingVisible}
+      />
+    </section>
+  );
+}

frontend-web/src/pages/official-task-analysis/TaskAnalysisPage.tsx

@@ -0,0 +1,460 @@
+/**
+ * 任务分析 · 列表页：对方系统推送过来的外部任务 + 我们的绑定状态。
+ * 任务卡片沿用多智能体定时任务的卡片视觉（浅红规范）。
+ */
+
+import {
+  CalendarClock,
+  CircleAlert,
+  CircleCheck,
+  Clock3,
+  GitBranch,
+  Link2,
+  Network,
+  Pencil,
+  Pause,
+  Play,
+  Plus,
+  Search,
+  Trash2,
+  Zap,
+} from "lucide-react";
+import { useMemo, useState } from "react";
+import { useNavigate } from "react-router-dom";
+import {
+  Button,
+  DialogPlugin,
+  Input,
+  Popconfirm,
+  Switch,
+  Tabs,
+  Tag,
+} from "tdesign-react";
+import { toast } from "sonner";
+
+import { OFFICIAL_DEERFLOW_ROUTE_PREFIX } from "@/core/official-deerflow/routes";
+
+import AgentFlow from "@/pages/official-multi-agent-tasks/components/AgentFlow";
+import type { MatSeat } from "@/pages/official-multi-agent-tasks/types";
+import {
+  formatDateTime,
+  formatRelative,
+} from "@/pages/official-multi-agent-tasks/utils";
+
+import BindingDialog from "./components/BindingDialog";
+import CreateTaskDialog, {
+  type TaCreateTaskPayload,
+} from "./components/CreateTaskDialog";
+import { chainStagesOf, useTaTaskActions, useTaTasks } from "./store";
+import type { TaTask, TaTaskViewLike } from "./page-types";
+import { taskBadgeOf } from "./badge";
+
+import "./task-analysis.css";
+
+type StatusFilter = "all" | "unbound" | "enabled" | "paused" | "failed";
+
+function isTaskLive(runs: TaTaskViewLike["runs"]) {
+  return runs.some((run) => run.status === "running" || run.status === "pending");
+}
+
+export default function TaskAnalysisPage() {
+  const navigate = useNavigate();
+  const { views, isLoading } = useTaTasks();
+  const actions = useTaTaskActions();
+  const [search, setSearch] = useState("");
+  const [filter, setFilter] = useState<StatusFilter>("all");
+  const [bindingTask, setBindingTask] = useState<TaTask | null>(null);
+  const [bindingVisible, setBindingVisible] = useState(false);
+  const [createVisible, setCreateVisible] = useState(false);
+  const [busyTaskId, setBusyTaskId] = useState<string | null>(null);
+
+  const visibleViews = useMemo(() => {
+    const keyword = search.trim().toLowerCase();
+    return views.filter(({ task, runs }) => {
+      const lastStatus = runs[0]?.status ?? null;
+      const bound = !!task.binding;
+      const live = isTaskLive(runs);
+      if (filter === "unbound" && bound) return false;
+      if (filter === "enabled" && (!bound || !task.binding!.enabled || live)) return false;
+      if (filter === "paused" && (!bound || task.binding!.enabled)) return false;
+      if (filter === "failed" && lastStatus !== "failed") return false;
+      if (!keyword) return true;
+      return (
+        task.name.toLowerCase().includes(keyword) ||
+        task.description.toLowerCase().includes(keyword) ||
+        task.sourceTaskId.toLowerCase().includes(keyword)
+      );
+    });
+  }, [views, search, filter]);
+
+  const stats = useMemo(
+    () => ({
+      total: views.length,
+      unbound: views.filter(({ task }) => !task.binding).length,
+      enabled: views.filter(({ task }) => task.binding?.enabled).length,
+      failed: views.filter(({ runs }) => runs[0]?.status === "failed").length,
+    }),
+    [views],
+  );
+
+  const openDetail = (taskId: string) => {
+    navigate(`${OFFICIAL_DEERFLOW_ROUTE_PREFIX}/workspace/task-analysis/${encodeURIComponent(taskId)}`);
+  };
+
+  const openBinding = (task: TaTask) => {
+    setBindingTask(task);
+    setBindingVisible(true);
+  };
+
+  const handleCreate = async (payload: TaCreateTaskPayload) => {
+    try {
+      const created = await actions.createTask.mutateAsync({
+        name: payload.name,
+        sourceTaskId: payload.sourceTaskId,
+        description: payload.description,
+      });
+      await actions.bind.mutateAsync({ taskId: created.id, payload: payload.binding });
+      toast.success("任务已创建并绑定分析方案");
+      setCreateVisible(false);
+    } catch {
+      toast.error("创建失败，请重试");
+    }
+  };
+
+  const handleBind = async (payload: Parameters<typeof actions.bind.mutateAsync>[0]["payload"]) => {
+    if (!bindingTask) return;
+    try {
+      await actions.bind.mutateAsync({ taskId: bindingTask.id, payload });
+      toast.success(bindingTask.binding ? "绑定已更新" : "分析方案已绑定");
+      setBindingVisible(false);
+      setBindingTask(null);
+    } catch {
+      toast.error("保存失败，请重试");
+    }
+  };
+
+  const runTrigger = async (taskId: string) => {
+    setBusyTaskId(taskId);
+    try {
+      await actions.triggerRun.mutateAsync(taskId);
+      toast.success("已触发执行");
+      openDetail(taskId);
+    } catch {
+      toast.error("触发失败：任务未绑定分析方案");
+    } finally {
+      setBusyTaskId(null);
+    }
+  };
+
+  const runToggle = async (task: TaTask) => {
+    setBusyTaskId(task.id);
+    try {
+      await actions.setEnabled.mutateAsync({ taskId: task.id, enabled: !task.binding?.enabled });
+      toast.success(task.binding?.enabled ? "已暂停" : "已恢复");
+    } catch {
+      toast.error("操作失败，请重试");
+    } finally {
+      setBusyTaskId(null);
+    }
+  };
+
+  const confirmDelete = (task: TaTask) => {
+    const dialog = DialogPlugin.confirm({
+      header: "删除外部任务",
+      body: `删除「${task.name}」将同时清除其绑定与全部执行记录（对应对外删除接口语义），确定删除？`,
+      confirmBtn: { content: "删除", theme: "danger" },
+      cancelBtn: "取消",
+      onConfirm: () => {
+        dialog.destroy();
+        void actions.removeTask.mutateAsync(task.id).then(() => {
+          toast.success("任务已删除");
+        });
+      },
+      onClose: () => dialog.destroy(),
+    });
+  };
+
+  return (
+    <section className="ta-page">
+      <header className="ta-page-head">
+        <div className="ta-page-head-copy">
+          <h1>任务分析定时</h1>
+          <p>
+            外部系统任务接入多智能体定时分析：绑定业务链与内容块映射，按计划执行并沉淀时间线版本。
+          </p>
+        </div>
+        <div className="ta-page-head-actions">
+          <Button
+            variant="outline"
+            onClick={() => {
+              const unbound = views.find(({ task }) => !task.binding)?.task;
+              if (unbound) openBinding(unbound);
+              else toast.info("当前任务均已绑定分析方案");
+            }}
+          >
+            配置未绑定任务
+          </Button>
+          <Button theme="primary" onClick={() => setCreateVisible(true)}>
+            <Plus aria-hidden="true" size={15} />
+            新建任务
+          </Button>
+        </div>
+      </header>
+
+      <div className="ta-stats" aria-label="任务统计">
+        <div className="ta-stat">
+          <span className="ta-stat-icon is-primary" aria-hidden="true">
+            <Network size={18} />
+          </span>
+          <div className="ta-stat-copy">
+            <strong>{stats.total}</strong>
+            <span>外部任务</span>
+          </div>
+        </div>
+        <div className="ta-stat">
+          <span className="ta-stat-icon is-warning" aria-hidden="true">
+            <Link2 size={18} />
+          </span>
+          <div className="ta-stat-copy">
+            <strong>{stats.unbound}</strong>
+            <span>未配置</span>
+          </div>
+        </div>
+        <div className="ta-stat">
+          <span className="ta-stat-icon is-success" aria-hidden="true">
+            <CircleCheck size={18} />
+          </span>
+          <div className="ta-stat-copy">
+            <strong>{stats.enabled}</strong>
+            <span>已启用</span>
+          </div>
+        </div>
+        <div className="ta-stat">
+          <span className="ta-stat-icon is-danger" aria-hidden="true">
+            <CircleAlert size={18} />
+          </span>
+          <div className="ta-stat-copy">
+            <strong>{stats.failed}</strong>
+            <span>最近失败</span>
+          </div>
+        </div>
+      </div>
+
+      <div className="ta-body">
+        <div className="ta-toolbar">
+          <div className="ta-toolbar-search">
+            <Input
+              clearable
+              placeholder="搜索任务名称 / 简介 / 对方任务号"
+              prefixIcon={<Search aria-hidden="true" size={15} />}
+              value={search}
+              onChange={(value) => setSearch(String(value ?? ""))}
+            />
+          </div>
+          <Tabs
+            className="ta-toolbar-tabs"
+            value={filter}
+            onChange={(value) => setFilter(value as StatusFilter)}
+          >
+            <Tabs.TabPanel label="全部" value="all" />
+            <Tabs.TabPanel label="未配置" value="unbound" />
+            <Tabs.TabPanel label="已启用" value="enabled" />
+            <Tabs.TabPanel label="已暂停" value="paused" />
+            <Tabs.TabPanel label="最近失败" value="failed" />
+          </Tabs>
+        </div>
+
+        {isLoading ? (
+          <div className="ta-state">
+            <Clock3 aria-hidden="true" className="ta-spin" size={18} />
+            正在加载外部任务…
+          </div>
+        ) : visibleViews.length ? (
+          <div className="ta-grid">
+            {visibleViews.map(({ task, runs }) => {
+              const badge = taskBadgeOf(task, runs);
+              const binding = task.binding;
+              const lastRun = runs[0] ?? null;
+              const busy = busyTaskId === task.id;
+              const stages = binding ? chainStagesOf(binding) : [];
+              const seats: MatSeat[] = stages
+                .flat()
+                .filter((id, index, all) => all.indexOf(id) === index)
+                .map((id) => ({
+                  agent_id: id,
+                  name: lastRun?.seatNames[id] ?? id,
+                }));
+              return (
+                <article className="ta-card" key={task.id}>
+                  <header className="ta-card-head" onClick={() => openDetail(task.id)}>
+                    <span className="ta-card-glyph" aria-hidden="true">
+                      <Network size={17} />
+                    </span>
+                    <div className="ta-card-title">
+                      <h3>
+                        {task.name}
+                        <em>{task.sourceTaskId}</em>
+                      </h3>
+                      <p>{task.description}</p>
+                    </div>
+                    <Tag size="small" theme={badge.theme} variant="light">
+                      {badge.label}
+                    </Tag>
+                  </header>
+
+                  {binding ? (
+                    <>
+                      <div className="ta-card-binding" onClick={() => openDetail(task.id)}>
+                        <span>
+                          <GitBranch aria-hidden="true" size={13} />
+                          {binding.chainTitle}
+                        </span>
+                        <span>
+                          <CalendarClock aria-hidden="true" size={13} />
+                          {binding.cronText}
+                        </span>
+                      </div>
+                      <div className="ta-card-flow" onClick={() => openDetail(task.id)}>
+                        <AgentFlow seats={seats} size="sm" stages={stages} />
+                      </div>
+                    </>
+                  ) : (
+                    <button className="ta-card-unbound" onClick={() => openBinding(task)} type="button">
+                      <GitBranch aria-hidden="true" size={15} />
+                      配置分析：选择业务链 · 内容块映射 · 执行频率
+                    </button>
+                  )}
+
+                  <footer className="ta-card-foot">
+                    <div className="ta-card-meta">
+                      <span>
+                        下次{" "}
+                        {binding?.enabled
+                          ? formatDateTime(nextRunAt(binding.cron))
+                          : binding
+                            ? "已暂停"
+                            : "未配置"}
+                      </span>
+                      <span>
+                        上次 {lastRun ? formatRelative(lastRun.startedAt) : "未执行"}
+                        {lastRun && lastRun.status !== "pending" && (
+                          <em className={`ta-run-text is-${lastRun.status}`}>
+                            · {lastRun.status === "succeeded" ? "成功" : lastRun.status === "failed" ? "失败" : "执行中"}
+                          </em>
+                        )}
+                      </span>
+                    </div>
+                    <div className="ta-card-ops">
+                      {binding && (
+                        <>
+                          <Button
+                            disabled={busy}
+                            loading={busy}
+                            size="small"
+                            theme="primary"
+                            variant="text"
+                            onClick={() => void runTrigger(task.id)}
+                          >
+                            <Zap aria-hidden="true" size={14} />
+                            立即执行
+                          </Button>
+                          <Button
+                            disabled={busy}
+                            size="small"
+                            variant="text"
+                            onClick={() => void runToggle(task)}
+                          >
+                            {binding.enabled ? (
+                              <Pause aria-hidden="true" size={14} />
+                            ) : (
+                              <Play aria-hidden="true" size={14} />
+                            )}
+                            {binding.enabled ? "暂停" : "恢复"}
+                          </Button>
+                          <Button size="small" variant="text" onClick={() => openBinding(task)}>
+                            <Pencil aria-hidden="true" size={14} />
+                            编辑绑定
+                          </Button>
+                        </>
+                      )}
+                      <Popconfirm
+                        content="删除后绑定与执行记录一并清除，确定？"
+                        placement="bottom-right"
+                        theme="danger"
+                        onConfirm={() => confirmDelete(task)}
+                      >
+                        <Button className="ta-danger-btn" size="small" variant="text">
+                          <Trash2 aria-hidden="true" size={14} />
+                        </Button>
+                      </Popconfirm>
+                    </div>
+                  </footer>
+                </article>
+              );
+            })}
+          </div>
+        ) : (
+          <div className="ta-state ta-state--empty">
+            <span className="ta-state-glyph" aria-hidden="true">
+              <Network size={22} />
+            </span>
+            <h3>{views.length ? "没有符合条件的任务" : "还没有外部任务"}</h3>
+            <p>
+              {views.length
+                ? "换个关键词或筛选条件试试。"
+                : "对方系统创建任务并调用创建接口后，任务会出现在这里。"}
+            </p>
+          </div>
+        )}
+
+        <p className="ta-footnote">
+          <CircleAlert aria-hidden="true" size={13} />
+          前端演示阶段：任务与执行数据保存在本浏览器，业务链来自真实接口；后端服务接入后自动切换。
+        </p>
+      </div>
+
+      <CreateTaskDialog
+        onClose={() => setCreateVisible(false)}
+        onSubmit={handleCreate}
+        submitting={actions.createTask.isPending}
+        visible={createVisible}
+      />
+
+      <BindingDialog
+        onClose={() => {
+          setBindingVisible(false);
+          setBindingTask(null);
+        }}
+        onSubmit={handleBind}
+        submitting={actions.bind.isPending}
+        task={bindingTask}
+        visible={bindingVisible}
+      />
+    </section>
+  );
+}
+
+/** 简易下次执行时间（覆盖本页生成的 cron 形态）。 */
+function nextRunAt(cron: string): string | null {
+  const parts = cron.trim().split(/\s+/);
+  if (parts.length !== 5) return null;
+  const [min, hour, , , wd] = parts;
+  const now = new Date();
+  const candidate = new Date(now);
+  const minute = parseInt(min ?? "0", 10);
+  const hourNum = parseInt(hour ?? "9", 10);
+  if (`${hour}` === "*") {
+    candidate.setMinutes(minute, 0, 0);
+    if (candidate <= now) candidate.setHours(candidate.getHours() + 1);
+  } else if (`${wd}` !== "*") {
+    const weekday = parseInt(wd ?? "1", 10);
+    candidate.setHours(hourNum, minute, 0, 0);
+    let delta = (weekday - candidate.getDay() + 7) % 7;
+    if (delta === 0 && candidate <= now) delta = 7;
+    candidate.setDate(candidate.getDate() + delta);
+  } else {
+    candidate.setHours(hourNum, minute, 0, 0);
+    if (candidate <= now) candidate.setDate(candidate.getDate() + 1);
+  }
+  return candidate.toISOString();
+}

frontend-web/src/pages/official-task-analysis/badge.ts

@@ -0,0 +1,19 @@
+import type { TaRunView, TaTask } from "./types";
+
+export interface TaBadge {
+  label: string;
+  theme: "success" | "warning" | "danger" | "default";
+}
+
+/** 任务卡右上角状态徽标：最近一次执行优先，其次启停，未绑定给引导态。 */
+export function taskBadgeOf(task: TaTask, runs: TaRunView[]): TaBadge {
+  const status = runs[0]?.status;
+  if (status === "running" || status === "pending") {
+    return { label: "执行中", theme: "warning" };
+  }
+  if (!task.binding) return { label: "未配置", theme: "default" };
+  if (!task.binding.enabled) return { label: "已暂停", theme: "default" };
+  if (status === "failed") return { label: "上次失败", theme: "danger" };
+  if (status === "succeeded") return { label: "运行正常", theme: "success" };
+  return { label: "已启用", theme: "success" };
+}

frontend-web/src/pages/official-task-analysis/components/BindingDialog.tsx

@@ -0,0 +1,223 @@
+/**
+ * 绑定配置弹窗：选业务链（真实 /api/roundtable-chains + 内置演示链）→ 配置
+ * 对方详情块 → 席位映射 → 执行频率 / 启用 / 总控提示。
+ */
+
+import { Radio } from "lucide-react";
+import { useEffect, useMemo, useState } from "react";
+import { useQuery } from "@tanstack/react-query";
+import { Button, Dialog, Switch, Textarea } from "tdesign-react";
+import { toast } from "sonner";
+
+import { listChains } from "@/roundtable-planning/api/chains";
+import type { BusinessChain } from "@/roundtable-planning/lib/business-chain";
+import { type ScheduleConfig } from "@/components/workspace/scheduled-tasks/schedule-builder";
+
+import AgentFlow from "@/pages/official-multi-agent-tasks/components/AgentFlow";
+import SchedulePicker from "@/pages/official-multi-agent-tasks/components/SchedulePicker";
+import type { MatSeat } from "@/pages/official-multi-agent-tasks/types";
+import {
+  defaultScheduleConfig,
+  describeScheduleConfig,
+} from "@/pages/official-multi-agent-tasks/utils";
+
+import type { TaBinding, TaBindingPayload, TaTask } from "../types";
+
+/** 内置演示链（种子里用它，保证无后端数据时链路可演示）。 */
+export const SEED_CHAIN: BusinessChain = {
+  id: "seed-chain-std",
+  title: "标准舆情分析链（内置）",
+  description: "采集 + 核查并行 → 研判 → 成稿，演示用内置编排。",
+  seats: [
+    { agent_id: "collector", name: "全网舆情采集" },
+    { agent_id: "verifier", name: "数据核查员" },
+    { agent_id: "analyst", name: "舆情研判分析" },
+    { agent_id: "writer", name: "报告撰写助手" },
+  ],
+  stages: [
+    ["collector", "verifier"],
+    ["analyst"],
+    ["writer"],
+  ],
+  stageGoals: null,
+  isPublic: true,
+};
+
+export function chainStages(chain: BusinessChain): string[][] {
+  if (chain.stages?.length) return chain.stages;
+  return chain.seats.map((seat) => [seat.agent_id]);
+}
+
+export function chainSeats(chain: BusinessChain): MatSeat[] {
+  return chain.seats.map((seat) => ({
+    agent_id: seat.agent_id,
+    name: seat.name,
+    description: seat.description,
+  }));
+}
+
+/** 模块级空数组：data 未就绪时保持引用稳定，避免 effect 依赖无限循环。 */
+export const NO_CHAINS: BusinessChain[] = [];
+
+export default function BindingDialog({
+  visible,
+  task,
+  submitting,
+  onClose,
+  onSubmit,
+}: {
+  visible: boolean;
+  task: TaTask | null;
+  submitting: boolean;
+  onClose: () => void;
+  onSubmit: (payload: TaBindingPayload) => void;
+}) {
+  const [chainId, setChainId] = useState("");
+  const [schedule, setSchedule] = useState<ScheduleConfig>(defaultScheduleConfig());
+  const [enabled, setEnabled] = useState(true);
+  const [coordinatorPrompt, setCoordinatorPrompt] = useState("");
+
+  const { data, isLoading: chainsLoading } = useQuery({
+    queryKey: ["ta", "chains"],
+    queryFn: async () => {
+      const [mine, publicChains] = await Promise.all([
+        listChains("mine").catch(() => [] as BusinessChain[]),
+        listChains("public").catch(() => [] as BusinessChain[]),
+      ]);
+      const byId = new Map<string, BusinessChain>();
+      for (const chain of [SEED_CHAIN, ...mine, ...publicChains]) {
+        if (!byId.has(chain.id)) byId.set(chain.id, chain);
+      }
+      return [...byId.values()];
+    },
+    enabled: visible,
+  });
+  const chains = data ?? NO_CHAINS;
+
+  useEffect(() => {
+    if (!visible) return;
+    const binding: TaBinding | null = task?.binding ?? null;
+    const initialChain =
+      chains.find((chain) => chain.id === binding?.chainId) ??
+      chains[0] ??
+      SEED_CHAIN;
+    setChainId(binding?.chainId ?? initialChain.id);
+    setEnabled(binding?.enabled ?? true);
+    setCoordinatorPrompt(binding?.coordinatorPrompt ?? "");
+  }, [visible, task, chains]);
+
+  const selectedChain = useMemo(
+    () => chains.find((chain) => chain.id === chainId) ?? null,
+    [chains, chainId],
+  );
+  const handleChainChange = (nextId: string) => {
+    setChainId(nextId);
+  };
+
+  const trySubmit = () => {
+    if (!selectedChain) {
+      toast.warning("请选择业务链");
+      return;
+    }
+    onSubmit({
+      chainId: selectedChain.id,
+      chainTitle: selectedChain.title,
+      schedule,
+      enabled,
+      coordinatorPrompt: coordinatorPrompt.trim() || undefined,
+    });
+  };
+
+  return (
+    <Dialog
+      className="ta-binding-dialog"
+      closeBtn
+      closeOnOverlayClick={false}
+      destroyOnClose
+      footer={
+        <div className="ta-dialog-footer">
+          <Button variant="outline" onClick={onClose}>
+            取消
+          </Button>
+          <Button loading={submitting} theme="primary" onClick={trySubmit}>
+            {task?.binding ? "保存绑定" : "创建绑定"}
+          </Button>
+        </div>
+      }
+      header={task?.binding ? `编辑绑定 · ${task.name}` : `配置分析 · ${task?.name ?? ""}`}
+      placement="top"
+      visible={visible}
+      width={860}
+      onClose={onClose}
+    >
+      <div className="ta-binding">
+        <section className="ta-binding-section">
+          <header className="ta-binding-section-head">
+            <h4>选择业务链</h4>
+            <span>定时执行将按业务链的多智能体关系（同阶段并行 · 阶段间串行）运行</span>
+          </header>
+          {chainsLoading ? (
+            <p className="ta-binding-tip">正在加载业务链…</p>
+          ) : (
+            <div className="ta-chain-grid">
+              {chains.map((chain) => {
+                const active = chain.id === chainId;
+                return (
+                  <button
+                    className={`ta-chain-card${active ? " is-active" : ""}`}
+                    key={chain.id}
+                    onClick={() => handleChainChange(chain.id)}
+                    type="button"
+                  >
+                    <header>
+                      <Radio aria-hidden="true" size={14} />
+                      <span className="ta-chain-card-title">{chain.title}</span>
+                      <em>
+                        {chain.seats.length} 智能体 · {chainStages(chain).length} 阶段
+                      </em>
+                    </header>
+                    {chain.description && <p>{chain.description}</p>}
+                    <div className="ta-chain-card-flow">
+                      <AgentFlow
+                        seats={chainSeats(chain)}
+                        size="sm"
+                        stages={chainStages(chain)}
+                      />
+                    </div>
+                  </button>
+                );
+              })}
+            </div>
+          )}
+        </section>
+
+        <section className="ta-binding-section">
+          <header className="ta-binding-section-head">
+            <h4>执行设置</h4>
+            <span>{describeScheduleConfig(schedule)} 自动执行</span>
+          </header>
+          <div className="ta-binding-settings">
+            <div className="ta-field">
+              <span className="ta-field-label">执行频率</span>
+              <SchedulePicker value={schedule} onChange={setSchedule} />
+            </div>
+            <label className="ta-field">
+              <span className="ta-field-label">总控提示（可选）</span>
+              <Textarea
+                autosize={{ minRows: 2, maxRows: 4 }}
+                maxlength={500}
+                placeholder="写给总控（协调者）的编排要求；问答沉淀的「方向备注」也会注入后续执行"
+                value={coordinatorPrompt}
+                onChange={(value) => setCoordinatorPrompt(String(value ?? ""))}
+              />
+            </label>
+            <div className="ta-field ta-field--switch">
+              <span className="ta-field-label">绑定后立即启用</span>
+              <Switch value={enabled} onChange={(value) => setEnabled(!!value)} />
+            </div>
+          </div>
+        </section>
+      </div>
+    </Dialog>
+  );
+}

frontend-web/src/pages/official-task-analysis/components/CreateTaskDialog.tsx

@@ -0,0 +1,242 @@
+/**
+ * 新建任务弹窗（一步式）：任务信息（名称/任务号/简介）+ 业务链选择 +
+ * 执行设置（频率 / 总控提示 / 启用），提交后直接创建任务并完成绑定。
+ */
+
+import { Radio } from "lucide-react";
+import { useEffect, useMemo, useState } from "react";
+import { useQuery } from "@tanstack/react-query";
+import { Button, Dialog, Input, Switch, Textarea } from "tdesign-react";
+import { toast } from "sonner";
+
+import { listChains } from "@/roundtable-planning/api/chains";
+import type { BusinessChain } from "@/roundtable-planning/lib/business-chain";
+import { type ScheduleConfig } from "@/components/workspace/scheduled-tasks/schedule-builder";
+
+import AgentFlow from "@/pages/official-multi-agent-tasks/components/AgentFlow";
+import SchedulePicker from "@/pages/official-multi-agent-tasks/components/SchedulePicker";
+import {
+  defaultScheduleConfig,
+  describeScheduleConfig,
+} from "@/pages/official-multi-agent-tasks/utils";
+import type { MatSeat } from "@/pages/official-multi-agent-tasks/types";
+
+import { NO_CHAINS, SEED_CHAIN, chainSeats, chainStages } from "./BindingDialog";
+import type { TaBindingPayload } from "../types";
+
+export interface TaCreateTaskPayload {
+  name: string;
+  sourceTaskId?: string;
+  description?: string;
+  binding: TaBindingPayload;
+}
+
+export default function CreateTaskDialog({
+  visible,
+  submitting,
+  onClose,
+  onSubmit,
+}: {
+  visible: boolean;
+  submitting: boolean;
+  onClose: () => void;
+  onSubmit: (payload: TaCreateTaskPayload) => void;
+}) {
+  const [name, setName] = useState("");
+  const [sourceTaskId, setSourceTaskId] = useState("");
+  const [description, setDescription] = useState("");
+  const [chainId, setChainId] = useState("");
+  const [schedule, setSchedule] = useState<ScheduleConfig>(defaultScheduleConfig());
+  const [enabled, setEnabled] = useState(true);
+  const [coordinatorPrompt, setCoordinatorPrompt] = useState("");
+
+  const { data, isLoading: chainsLoading } = useQuery({
+    queryKey: ["ta", "chains"],
+    queryFn: async () => {
+      const [mine, publicChains] = await Promise.all([
+        listChains("mine").catch(() => [] as BusinessChain[]),
+        listChains("public").catch(() => [] as BusinessChain[]),
+      ]);
+      const byId = new Map<string, BusinessChain>();
+      for (const chain of [SEED_CHAIN, ...mine, ...publicChains]) {
+        if (!byId.has(chain.id)) byId.set(chain.id, chain);
+      }
+      return [...byId.values()];
+    },
+    enabled: visible,
+  });
+  const chains = data ?? NO_CHAINS;
+
+  useEffect(() => {
+    if (!visible) return;
+    setName("");
+    setSourceTaskId("");
+    setDescription("");
+    setSchedule(defaultScheduleConfig());
+    setEnabled(true);
+    setCoordinatorPrompt("");
+    setChainId((prev) => (chains.some((chain) => chain.id === prev) ? prev : chains[0]?.id ?? SEED_CHAIN.id));
+  }, [visible, chains]);
+
+  const selectedChain = useMemo(
+    () => chains.find((chain) => chain.id === chainId) ?? null,
+    [chains, chainId],
+  );
+
+  const trySubmit = () => {
+    if (!name.trim()) {
+      toast.warning("请填写任务名称");
+      return;
+    }
+    if (!selectedChain) {
+      toast.warning("请选择业务链");
+      return;
+    }
+    onSubmit({
+      name,
+      sourceTaskId: sourceTaskId.trim() || undefined,
+      description: description.trim() || undefined,
+      binding: {
+        chainId: selectedChain.id,
+        chainTitle: selectedChain.title,
+        schedule,
+        enabled,
+        coordinatorPrompt: coordinatorPrompt.trim() || undefined,
+      },
+    });
+  };
+
+  return (
+    <Dialog
+      className="ta-create-dialog"
+      closeBtn
+      closeOnOverlayClick={false}
+      destroyOnClose
+      footer={
+        <div className="ta-dialog-footer">
+          <Button variant="outline" onClick={onClose}>
+            取消
+          </Button>
+          <Button loading={submitting} theme="primary" onClick={trySubmit}>
+            创建任务并绑定
+          </Button>
+        </div>
+      }
+      header="新建任务"
+      placement="top"
+      visible={visible}
+      width={860}
+      onClose={onClose}
+    >
+      <div className="ta-binding">
+        <section className="ta-binding-section">
+          <header className="ta-binding-section-head">
+            <h4>任务信息</h4>
+            <span>由本页录入；接入对方系统后，推送创建的任务同样出现在列表中</span>
+          </header>
+          <div className="ta-create-grid">
+            <label className="ta-field">
+              <span className="ta-field-label">
+                任务名称<i className="ta-required">*</i>
+              </span>
+              <Input
+                clearable
+                maxlength={40}
+                placeholder="例如：东南亚舆论环境监测"
+                value={name}
+                onChange={(value) => setName(String(value ?? ""))}
+              />
+            </label>
+            <label className="ta-field">
+              <span className="ta-field-label">对方任务号（可选）</span>
+              <Input
+                clearable
+                maxlength={32}
+                placeholder="留空自动生成"
+                value={sourceTaskId}
+                onChange={(value) => setSourceTaskId(String(value ?? ""))}
+              />
+            </label>
+            <label className="ta-field ta-field--wide">
+              <span className="ta-field-label">任务简介</span>
+              <Textarea
+                autosize={{ minRows: 2, maxRows: 4 }}
+                maxlength={200}
+                placeholder="一句话说明任务背景与分析目标，定时执行时作为任务上下文注入各智能体"
+                value={description}
+                onChange={(value) => setDescription(String(value ?? ""))}
+              />
+            </label>
+          </div>
+        </section>
+
+        <section className="ta-binding-section">
+          <header className="ta-binding-section-head">
+            <h4>执行设置</h4>
+            <span>{describeScheduleConfig(schedule)} 自动执行</span>
+          </header>
+          <div className="ta-binding-settings">
+            <div className="ta-field">
+              <span className="ta-field-label">执行频率</span>
+              <SchedulePicker value={schedule} onChange={setSchedule} />
+            </div>
+            <label className="ta-field">
+              <span className="ta-field-label">总控提示（可选）</span>
+              <Textarea
+                autosize={{ minRows: 2, maxRows: 4 }}
+                maxlength={500}
+                placeholder="写给总控（协调者）的编排要求；问答沉淀的「方向备注」也会注入后续执行"
+                value={coordinatorPrompt}
+                onChange={(value) => setCoordinatorPrompt(String(value ?? ""))}
+              />
+            </label>
+            <div className="ta-field ta-field--switch">
+              <span className="ta-field-label">创建后立即启用</span>
+              <Switch value={enabled} onChange={(value) => setEnabled(!!value)} />
+            </div>
+          </div>
+        </section>
+        <section className="ta-binding-section">
+          <header className="ta-binding-section-head">
+            <h4>选择业务链</h4>
+            <span>定时执行按业务链的多智能体关系（同阶段并行 · 阶段间串行）运行</span>
+          </header>
+          {chainsLoading ? (
+            <p className="ta-binding-tip">正在加载业务链…</p>
+          ) : (
+            <div className="ta-chain-grid">
+              {chains.map((chain) => {
+                const active = chain.id === chainId;
+                return (
+                  <button
+                    className={`ta-chain-card${active ? " is-active" : ""}`}
+                    key={chain.id}
+                    onClick={() => setChainId(chain.id)}
+                    type="button"
+                  >
+                    <header>
+                      <Radio aria-hidden="true" size={14} />
+                      <span className="ta-chain-card-title">{chain.title}</span>
+                      <em>
+                        {chain.seats.length} 智能体 · {chainStages(chain).length} 阶段
+                      </em>
+                    </header>
+                    {chain.description && <p>{chain.description}</p>}
+                    <div className="ta-chain-card-flow">
+                      <AgentFlow
+                        seats={chainSeats(chain) as MatSeat[]}
+                        size="sm"
+                        stages={chainStages(chain)}
+                      />
+                    </div>
+                  </button>
+                );
+              })}
+            </div>
+          )}
+        </section>
+
+      </div>
+    </Dialog>
+  );
+}

frontend-web/src/pages/official-task-analysis/components/SeatProcessDrawer.tsx

@@ -0,0 +1,122 @@
+/**
+ * 席位执行过程抽屉：点击流程图节点打开，展示该智能体在某次运行里的完整过程
+ * （输入 → AI 分析 → 工具调用 → 产物块）。后端就绪后由 runs/{id}/messages?seat= 提供数据。
+ */
+
+import { ArrowDownRight, Bot, Wrench } from "lucide-react";
+import { Drawer, Tag } from "tdesign-react";
+
+import { seatColorIndex } from "@/pages/official-multi-agent-tasks/utils";
+import { buildSeatProcess } from "../store";
+import type { TaRunView, TaSeatState, TaTask } from "../types";
+
+const STATE_THEME: Record<TaSeatState, "success" | "warning" | "danger" | "default"> = {
+  done: "success",
+  running: "warning",
+  failed: "danger",
+  waiting: "default",
+};
+
+const STATE_TEXT: Record<TaSeatState, string> = {
+  waiting: "等待执行",
+  running: "执行中",
+  done: "已完成",
+  failed: "执行失败",
+};
+
+const KIND_LABEL: Record<string, string> = {
+  input: "输入",
+  ai: "分析",
+  tool: "工具",
+  artifact: "产物",
+};
+
+function formatDuration(ms: number): string {
+  if (!Number.isFinite(ms) || ms <= 0) return "—";
+  const seconds = Math.round(ms / 1000);
+  if (seconds < 60) return `${seconds} 秒`;
+  return `${Math.floor(seconds / 60)} 分 ${seconds % 60} 秒`;
+}
+
+export default function SeatProcessDrawer({
+  run,
+  seatAgentId,
+  task,
+  onClose,
+}: {
+  run: TaRunView | null;
+  seatAgentId: string | null;
+  task: TaTask | undefined;
+  onClose: () => void;
+}) {
+  const visible = !!run && !!seatAgentId;
+  const process =
+    visible && run ? buildSeatProcess(run, seatAgentId!, task) : null;
+  const info = run?.seatInfos[seatAgentId ?? ""] ?? null;
+
+  return (
+    <Drawer
+      className="ta-seat-drawer"
+      closeBtn
+      destroyOnClose
+      footer={null}
+      header={
+        process ? (
+          <div className="ta-seat-header">
+            <span className={`ta-seat-glyph ta-seat--c${seatColorIndex(seatAgentId ?? "")}`} aria-hidden="true">
+              <Bot size={18} />
+            </span>
+            <div className="ta-seat-heading">
+              <h3>{process.seatName}</h3>
+              <p>执行过程 · {run?.trigger === "manual" ? "手动执行（总控问答）" : "定时执行"}</p>
+            </div>
+            {info && (
+              <Tag size="small" theme={STATE_THEME[info.state]} variant="light">
+                {STATE_TEXT[info.state]}
+              </Tag>
+            )}
+          </div>
+        ) : null
+      }
+      placement="right"
+      size="520px"
+      visible={visible}
+      onClose={onClose}
+    >
+      {process && (
+        <div className="ta-seat-process">
+          {info && (
+            <div className="ta-seat-meta">
+              <span>状态：{STATE_TEXT[info.state]}</span>
+              <span>耗时：{formatDuration(info.durationMs)}</span>
+              <span>所属阶段：第 {(run?.stages.findIndex((stage) => stage.includes(seatAgentId ?? "")) ?? -1) + 1} 阶段</span>
+            </div>
+          )}
+          <ol className="ta-seat-timeline">
+            {process.messages.map((message, index) => (
+              <li className={`ta-seat-item is-${message.kind}`} key={index}>
+                <span className="ta-seat-item-marker">
+                  {message.kind === "tool" ? (
+                    <Wrench size={11} />
+                  ) : (
+                    <ArrowDownRight size={11} />
+                  )}
+                </span>
+                <div className="ta-seat-item-body">
+                  <header>
+                    <em className={`ta-seat-item-kind is-${message.kind}`}>
+                      {KIND_LABEL[message.kind] ?? message.kind}
+                    </em>
+                    <span className="ta-seat-item-title">{message.title}</span>
+                  </header>
+                  <p>{message.content}</p>
+                  {message.detail && <pre>{message.detail}</pre>}
+                </div>
+              </li>
+            ))}
+          </ol>
+        </div>
+      )}
+    </Drawer>
+  );
+}

frontend-web/src/pages/official-task-analysis/page-types.ts

@@ -0,0 +1,4 @@
+import type { TaRunView, TaTask } from "./types";
+
+export type { TaRunView, TaTask };
+export type TaTaskViewLike = { task: TaTask; runs: TaRunView[] };

frontend-web/src/pages/official-task-analysis/store.ts

@@ -0,0 +1,611 @@
+/**
+ * 外部任务定时分析的本地 mock 存储层（与 multi-agent-tasks 同款模式）。
+ *
+ * - 任务/绑定/运行记录落 localStorage；「立即执行 / 模拟手动执行」写入运行记录，
+ *   状态由 startedAt + durationMs 按当前时间推导（排队 → 按阶段推进 → 成功/失败），
+ *   配合 React Query 轮询获得无后端的"实时"流程图推进。
+ * - 席位执行过程消息在读取侧按席位生成（buildSeatProcess），不入库。
+ * - 对外接口（创建/删除/详情时间线/问答）就绪后替换本层。
+ */
+
+import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query";
+
+import { buildCron } from "@/components/workspace/scheduled-tasks/schedule-builder";
+import { describeScheduleConfig } from "@/pages/official-multi-agent-tasks/utils";
+import { uuid } from "@/core/utils/uuid";
+
+import type {
+  TaBlockMapping,
+  TaBinding,
+  TaBindingPayload,
+  TaRunRecord,
+  TaRunStatus,
+  TaRunView,
+  TaSeatRunInfo,
+  TaSeatState,
+  TaTask,
+} from "./types";
+
+const STORAGE_KEY = "deerflow.official.task-analysis.v2";
+
+interface TaStore {
+  tasks: TaTask[];
+  runs: TaRunRecord[];
+  /** 上次绑定时选择的业务链 id（创建接口响应里的 default_chain_id 记忆）。 */
+  lastChainId: string | null;
+}
+
+const PENDING_MS = 2_000;
+const STAGE_MS = 6_000;
+
+/** 对方系统定义的详情块类型（mock 固定清单；接接口后由创建方传入）。 */
+export const TA_BLOCK_TYPES: Array<{ key: string; name: string }> = [
+  { key: "background", name: "任务背景" },
+  { key: "dynamics", name: "动态分析" },
+  { key: "risk", name: "风险研判" },
+  { key: "action", name: "行动建议" },
+  { key: "conclusion", name: "结论汇总" },
+];
+
+function readStore(): TaStore {
+  try {
+    const raw = window.localStorage.getItem(STORAGE_KEY);
+    if (raw) {
+      const parsed = JSON.parse(raw) as Partial<TaStore>;
+      if (Array.isArray(parsed.tasks)) {
+        return {
+          tasks: parsed.tasks as TaTask[],
+          runs: parsed.runs ?? [],
+          lastChainId: parsed.lastChainId ?? null,
+        };
+      }
+    }
+  } catch {
+    // 解析失败按首次访问处理。
+  }
+  const seeded = seedStore();
+  writeStore(seeded);
+  return seeded;
+}
+
+function writeStore(store: TaStore) {
+  try {
+    window.localStorage.setItem(STORAGE_KEY, JSON.stringify(store));
+  } catch {
+    // 存储被禁用时静默降级为内存态。
+  }
+}
+
+function mutateStore(mutator: (store: TaStore) => TaStore): TaStore {
+  const next = mutator(readStore());
+  writeStore(next);
+  return next;
+}
+
+function hoursAgo(hours: number): string {
+  return new Date(Date.now() - hours * 3_600_000).toISOString();
+}
+
+function seedStore(): TaStore {
+  const seats = {
+    collector: "全网舆情采集",
+    analyst: "舆情研判分析",
+    writer: "报告撰写助手",
+    verifier: "数据核查员",
+    forecaster: "趋势预测专家",
+  };
+
+  const tasks: TaTask[] = [
+    {
+      id: "ta-task-1",
+      sourceTaskId: "TS-2026-0147",
+      name: "东南亚舆论环境监测",
+      description: "对方系统推送的任务：持续跟踪东南亚重点渠道舆论动向，输出结构化分析详情。",
+      binding: {
+        chainId: "seed-chain-std",
+        chainTitle: "标准舆情分析链",
+        cron: "0 9 * * *",
+        cronText: "每天 09:00",
+        enabled: true,
+        blocks: [
+          { key: "background", name: "任务背景", seatAgentId: "collector" },
+          { key: "dynamics", name: "动态分析", seatAgentId: "collector" },
+          { key: "risk", name: "风险研判", seatAgentId: "analyst" },
+          { key: "action", name: "行动建议", seatAgentId: "analyst" },
+          { key: "conclusion", name: "结论汇总", seatAgentId: "writer" },
+        ],
+        coordinatorPrompt: "采集阶段允许独立检索；研判阶段需引用采集阶段的原始线索。",
+        directionNotes: ["重点关注宗教节日前后的舆论波动。"],
+        timezone: "Asia/Shanghai",
+      },
+      createdAt: hoursAgo(24 * 12),
+    },
+    {
+      id: "ta-task-2",
+      sourceTaskId: "TS-2026-0152",
+      name: "竞品新品发布追踪",
+      description: "追踪 5 家核心竞品的新品发布与市场反应，按块更新详情内容。",
+      binding: {
+        chainId: "seed-chain-single",
+        chainTitle: "单阶段并行分析链",
+        cron: "30 8 * * 1",
+        cronText: "每周一 08:30",
+        enabled: false,
+        blocks: [
+          { key: "background", name: "任务背景", seatAgentId: "verifier" },
+          { key: "dynamics", name: "动态分析", seatAgentId: "collector" },
+          { key: "conclusion", name: "结论汇总", seatAgentId: "writer" },
+        ],
+        directionNotes: [],
+        timezone: "Asia/Shanghai",
+      },
+      createdAt: hoursAgo(24 * 8),
+    },
+    {
+      id: "ta-task-3",
+      sourceTaskId: "TS-2026-0160",
+      name: "行业政策周报编制",
+      description: "对方系统刚创建的任务，尚未配置分析方案。",
+      binding: null,
+      createdAt: hoursAgo(5),
+    },
+  ];
+
+  const stdStages = [
+    ["collector", "verifier"],
+    ["analyst"],
+    ["writer"],
+  ];
+  const singleStages = [["collector", "verifier", "analyst"]];
+  const stdNames = seats;
+
+  const runs: TaRunRecord[] = [
+    {
+      id: uuid() as string,
+      taskId: "ta-task-1",
+      trigger: "schedule",
+      scheduledFor: hoursAgo(2.1),
+      startedAt: hoursAgo(2.1),
+      durationMs: PENDING_MS + stdStages.length * STAGE_MS,
+      stages: stdStages,
+      stageGoals: ["并行完成全网采集与数据核查", "研判风险与行动方向", "汇总各块结论"],
+      seatNames: stdNames,
+      failStage: null,
+    },
+    {
+      id: uuid() as string,
+      taskId: "ta-task-1",
+      trigger: "manual",
+      scheduledFor: hoursAgo(20),
+      startedAt: hoursAgo(20),
+      durationMs: PENDING_MS + stdStages.length * STAGE_MS,
+      stages: stdStages,
+      stageGoals: ["并行完成全网采集与数据核查", "研判风险与行动方向", "汇总各块结论"],
+      seatNames: stdNames,
+      failStage: null,
+      manualSummary: "总控问答：要求研判阶段补充跨平台对比维度，已确认应用到后续执行。",
+    },
+    {
+      id: uuid() as string,
+      taskId: "ta-task-1",
+      trigger: "schedule",
+      scheduledFor: hoursAgo(26),
+      startedAt: hoursAgo(26),
+      durationMs: PENDING_MS + stdStages.length * STAGE_MS,
+      stages: stdStages,
+      stageGoals: ["并行完成全网采集与数据核查", "研判风险与行动方向", "汇总各块结论"],
+      seatNames: stdNames,
+      failStage: 1,
+    },
+    {
+      id: uuid() as string,
+      taskId: "ta-task-2",
+      trigger: "schedule",
+      scheduledFor: hoursAgo(24 * 3),
+      startedAt: hoursAgo(24 * 3),
+      durationMs: PENDING_MS + singleStages.length * STAGE_MS,
+      stages: singleStages,
+      stageGoals: ["三席并行：采集 / 核查 / 研判同时产出"],
+      seatNames: stdNames,
+      failStage: null,
+    },
+  ];
+
+  return { tasks, runs, lastChainId: "seed-chain-std" };
+}
+
+/** 由记录 + 当前时间推导运行视图：各席位状态/耗时、块产出状态。 */
+function resolveRunView(run: TaRunRecord, task: TaTask | undefined, now = Date.now()): TaRunView {
+  const startedAt = new Date(run.startedAt).getTime();
+  const elapsed = now - startedAt;
+  const stageTotal = Math.max(1, run.stages.length);
+  const duration = run.durationMs || PENDING_MS + stageTotal * STAGE_MS;
+
+  const buildSeatInfos = (
+    status: TaRunStatus,
+    activeStage: number,
+  ): Record<string, TaSeatRunInfo> => {
+    const infos: Record<string, TaSeatRunInfo> = {};
+    run.stages.forEach((stage, index) => {
+      const stageStart = PENDING_MS + index * STAGE_MS;
+      const stageEnd = stageStart + STAGE_MS;
+      const finished = status === "succeeded" || (status === "failed" && index < activeStage);
+      const active = !finished && index === activeStage && status !== "pending";
+      const state: TaSeatState = finished
+        ? "done"
+        : active
+          ? status === "failed"
+            ? "failed"
+            : "running"
+          : "waiting";
+      const durationMs = finished
+        ? Math.min(elapsed - stageStart, STAGE_MS)
+        : active
+          ? Math.min(elapsed - stageStart, STAGE_MS)
+          : 0;
+      stage.forEach((agentId) => {
+        infos[agentId] = { state, durationMs };
+      });
+    });
+    return infos;
+  };
+
+  let status: TaRunStatus;
+  let stageIndex: number;
+  let finishedAt: string | null = null;
+  let error: string | null = null;
+
+  if (elapsed < PENDING_MS) {
+    status = "pending";
+    stageIndex = 0;
+  } else {
+    const progress = Math.min(elapsed - PENDING_MS, duration - PENDING_MS);
+    stageIndex = Math.min(Math.floor(progress / STAGE_MS), stageTotal - 1);
+    if (run.failStage !== null && stageIndex >= run.failStage) {
+      status = "failed";
+      stageIndex = run.failStage;
+      finishedAt = new Date(startedAt + PENDING_MS + (run.failStage + 1) * STAGE_MS).toISOString();
+      const failSeat = run.stages[run.failStage]?.[0];
+      error = `阶段 ${run.failStage + 1} 执行失败：${run.seatNames[failSeat ?? ""] ?? "席位"} 模拟执行异常`;
+    } else if (elapsed >= duration) {
+      status = "succeeded";
+      stageIndex = stageTotal - 1;
+      finishedAt = new Date(startedAt + duration).toISOString();
+    } else {
+      status = "running";
+    }
+  }
+
+  const seatInfos = buildSeatInfos(status, status === "pending" ? -1 : stageIndex);
+
+  const blockStates = (task?.binding?.blocks ?? []).map((block) => ({
+    key: block.key,
+    name: block.name,
+    state: seatInfos[block.seatAgentId]?.state ?? "waiting",
+  }));
+
+  return {
+    ...run,
+    finishedAt,
+    status,
+    stageIndex,
+    stageTotal,
+    seatInfos,
+    error,
+    blockStates,
+  };
+}
+
+function buildTaskViews(now = Date.now()): Array<{ task: TaTask; runs: TaRunView[] }> {
+  const store = readStore();
+  return store.tasks.map((task) => {
+    const runs = store.runs
+      .filter((run) => run.taskId === task.id)
+      .map((run) => resolveRunView(run, task, now))
+      .sort((a, b) => (a.startedAt < b.startedAt ? 1 : -1));
+    return { task, runs };
+  });
+}
+
+function anyRunLive(views: Array<{ runs: TaRunView[] }> | undefined): boolean {
+  return !!views?.some(({ runs }) =>
+    runs.some((run) => run.status === "pending" || run.status === "running"),
+  );
+}
+
+/** 列表 + 每任务运行视图；运行中 5 秒轮询（设计定稿：流程图每 5 秒刷新）。 */
+export function useTaTasks() {
+  const query = useQuery({
+    queryKey: ["ta", "tasks"],
+    queryFn: () => buildTaskViews(),
+    refetchInterval: (q) =>
+      anyRunLive(q.state.data as Array<{ runs: TaRunView[] }> | undefined) ? 5_000 : false,
+  });
+  const views = query.data ?? [];
+  return { views, tasks: views.map((view) => view.task), isLoading: query.isLoading };
+}
+
+export function useTaTaskRuns(taskId: string | null) {
+  const query = useQuery({
+    queryKey: ["ta", "tasks", taskId],
+    queryFn: () => {
+      const view = buildTaskViews().find((item) => item.task.id === taskId);
+      return view?.runs ?? [];
+    },
+    enabled: !!taskId,
+    refetchInterval: (q) => {
+      const runs = q.state.data as TaRunView[] | undefined;
+      const live = !!runs?.some((run) => run.status === "pending" || run.status === "running");
+      return live ? 5_000 : false;
+    },
+  });
+  return { runs: query.data ?? [], isLoading: query.isLoading };
+}
+
+export function useTaTaskActions() {
+  const queryClient = useQueryClient();
+  const invalidate = () => queryClient.invalidateQueries({ queryKey: ["ta"] });
+
+  /** 页面手动创建外部任务（名称必填；对方任务号留空则自动生成）。 */
+  const createTask = useMutation({
+    mutationFn: async (payload: {
+      name: string;
+      sourceTaskId?: string;
+      description?: string;
+    }) => {
+      const task: TaTask = {
+        id: uuid() as string,
+        sourceTaskId:
+          payload.sourceTaskId?.trim() ||
+          `TS-${new Date().getFullYear()}-${String(Date.now()).slice(-5)}`,
+        name: payload.name.trim(),
+        description: payload.description?.trim() || "",
+        binding: null,
+        createdAt: new Date().toISOString(),
+      };
+      mutateStore((store) => ({ ...store, tasks: [task, ...store.tasks] }));
+      return task;
+    },
+    onSuccess: invalidate,
+  });
+
+  /** 绑定 / 修改绑定（记录 lastChainId 供创建接口返回默认项）。 */
+  const bind = useMutation({
+    mutationFn: async ({ taskId, payload }: { taskId: string; payload: TaBindingPayload }) => {
+      const cron = buildCron(payload.schedule);
+      mutateStore((store) => ({
+        ...store,
+        lastChainId: payload.chainId,
+        tasks: store.tasks.map((task) =>
+          task.id === taskId
+            ? {
+                ...task,
+                binding: {
+                  chainId: payload.chainId,
+                  chainTitle: payload.chainTitle,
+                  cron,
+                  cronText: describeScheduleConfig(payload.schedule),
+                  enabled: payload.enabled,
+                  blocks: payload.blocks ?? task.binding?.blocks ?? [],
+                  coordinatorPrompt: payload.coordinatorPrompt,
+                  directionNotes: task.binding?.directionNotes ?? [],
+                  timezone: task.binding?.timezone ?? "Asia/Shanghai",
+                } satisfies TaBinding,
+              }
+            : task,
+        ),
+      }));
+    },
+    onSuccess: invalidate,
+  });
+
+  const unbind = useMutation({
+    mutationFn: async (taskId: string) => {
+      mutateStore((store) => ({
+        ...store,
+        tasks: store.tasks.map((task) =>
+          task.id === taskId ? { ...task, binding: null } : task,
+        ),
+        runs: store.runs.filter((run) => run.taskId !== taskId),
+      }));
+    },
+    onSuccess: invalidate,
+  });
+
+  const setEnabled = useMutation({
+    mutationFn: async ({ taskId, enabled }: { taskId: string; enabled: boolean }) => {
+      mutateStore((store) => ({
+        ...store,
+        tasks: store.tasks.map((task) =>
+          task.id === taskId && task.binding
+            ? { ...task, binding: { ...task.binding, enabled } }
+            : task,
+        ),
+      }));
+    },
+    onSuccess: invalidate,
+  });
+
+  /** 删除外部任务（对应对外删除接口：任务 + 绑定 + 执行记录彻底删除）。 */
+  const removeTask = useMutation({
+    mutationFn: async (taskId: string) => {
+      mutateStore((store) => ({
+        ...store,
+        tasks: store.tasks.filter((task) => task.id !== taskId),
+        runs: store.runs.filter((run) => run.taskId !== taskId),
+      }));
+    },
+    onSuccess: invalidate,
+  });
+
+  /** 立即触发一次定时执行（schedule）。 */
+  const triggerRun = useMutation({
+    mutationFn: async (taskId: string) => {
+      const store = readStore();
+      const task = store.tasks.find((item) => item.id === taskId);
+      const binding = task?.binding;
+      if (!task || !binding) throw new Error("任务未绑定分析方案");
+      const stages = chainStagesOf(binding, store);
+      const now = new Date().toISOString();
+      const run: TaRunRecord = {
+        id: uuid() as string,
+        taskId,
+        trigger: "schedule",
+        scheduledFor: now,
+        startedAt: now,
+        durationMs: PENDING_MS + stages.length * STAGE_MS,
+        stages,
+        stageGoals: stages.map((_, index) => stageGoalOf(index)),
+        seatNames: Object.fromEntries(stages.flat().map((id) => [id, seatDisplayName(id)])),
+        failStage: null,
+      };
+      mutateStore((current) => ({ ...current, runs: [run, ...current.runs] }));
+      return run;
+    },
+    onSuccess: invalidate,
+  });
+
+  /** 模拟一次手动执行（对方详情页总控问答产生）。 */
+  const triggerManualRun = useMutation({
+    mutationFn: async (taskId: string) => {
+      const store = readStore();
+      const task = store.tasks.find((item) => item.id === taskId);
+      const binding = task?.binding;
+      if (!task || !binding) throw new Error("任务未绑定分析方案");
+      const stages = chainStagesOf(binding, store);
+      const now = new Date().toISOString();
+      const run: TaRunRecord = {
+        id: uuid() as string,
+        taskId,
+        trigger: "manual",
+        scheduledFor: now,
+        startedAt: now,
+        durationMs: PENDING_MS + stages.length * STAGE_MS,
+        stages,
+        stageGoals: stages.map((_, index) => stageGoalOf(index)),
+        seatNames: Object.fromEntries(stages.flat().map((id) => [id, seatDisplayName(id)])),
+        failStage: null,
+        manualSummary: "总控问答：按对方详情页提问，总控给对应席位派活并汇总结论。",
+      };
+      mutateStore((current) => ({ ...current, runs: [run, ...current.runs] }));
+      return run;
+    },
+    onSuccess: invalidate,
+  });
+
+  return { createTask, bind, unbind, setEnabled, removeTask, triggerRun, triggerManualRun };
+}
+
+/** 绑定链的编排（种子任务用内置编排；真实业务链按其 stages 展开，null = 线性）。 */
+export function chainStagesOf(
+  binding: Pick<TaBinding, "chainId">,
+  store: TaStore = readStore(),
+): string[][] {
+  if (binding.chainId === "seed-chain-std") {
+    return [
+      ["collector", "verifier"],
+      ["analyst"],
+      ["writer"],
+    ];
+  }
+  if (binding.chainId === "seed-chain-single") {
+    return [["collector", "verifier", "analyst"]];
+  }
+  // 真实业务链的编排由页面侧传入的 chain 数据展开；存储层兜底为单席位串行。
+  const seats = seedSeatsOf(store);
+  return seats.map((agentId) => [agentId]);
+}
+
+function seedSeatsOf(_store: TaStore): string[] {
+  return ["collector", "verifier", "analyst", "writer"];
+}
+
+/** 演示用的阶段目标（seed 链固定；真实链由绑定快照提供）。 */
+function stageGoalOf(stageIndex: number): string | null {
+  const goals = ["并行完成全网采集与数据核查", "研判风险与行动方向", "汇总各块结论"];
+  return goals[stageIndex] ?? null;
+}
+
+export function seatDisplayName(agentId: string): string {
+  const names: Record<string, string> = {
+    collector: "全网舆情采集",
+    verifier: "数据核查员",
+    analyst: "舆情研判分析",
+    writer: "报告撰写助手",
+    forecaster: "趋势预测专家",
+  };
+  return names[agentId] ?? agentId;
+}
+
+// ── 席位执行过程（读取侧生成 mock）────────────────────────────────────────
+
+export interface TaSeatMessage {
+  kind: "input" | "ai" | "tool" | "artifact";
+  title: string;
+  content: string;
+  detail?: string;
+}
+
+/** 生成某席位在某次运行里的执行过程消息（后端就绪后换 runs/{id}/messages?seat=）。 */
+export function buildSeatProcess(
+  run: TaRunView,
+  seatAgentId: string,
+  task: TaTask | undefined,
+): { seatName: string; messages: TaSeatMessage[]; blocks: TaBlockMapping[] } {
+  const seatName = run.seatNames[seatAgentId] ?? seatAgentId;
+  const info = run.seatInfos[seatAgentId] ?? { state: "waiting" as TaSeatState, durationMs: 0 };
+  const stageIndex = run.stages.findIndex((stage) => stage.includes(seatAgentId));
+  const blocks = (task?.binding?.blocks ?? []).filter(
+    (block) => block.seatAgentId === seatAgentId,
+  );
+
+  const messages: TaSeatMessage[] = [
+    {
+      kind: "input",
+      title: "输入 · 任务上下文",
+      content:
+        `${task?.name ?? ""}（对方任务 ${task?.sourceTaskId ?? "-"}）\n` +
+        `${task?.description ?? ""}\n\n` +
+        (blocks.length
+          ? `本席位负责产出：${blocks.map((block) => block.name).join("、")}。`
+          : "按业务链席位职责执行。") +
+        (stageIndex > 0
+          ? `\n上一阶段产出摘要：${run.stages[stageIndex - 1]
+              ?.map((id) => run.seatNames[id])
+              .join("、")} 的中间结论已注入。`
+          : ""),
+    },
+  ];
+
+  if (info.state === "waiting") {
+    return { seatName, messages, blocks };
+  }
+
+  messages.push({
+    kind: "ai",
+    title: "AI 分析",
+    content:
+      `（${seatName} 的分析过程 · 演示内容）\n` +
+      `围绕任务简介拆解检索方向，交叉核对来源可信度，形成本席位负责块的结构化结论。`,
+  });
+
+  messages.push({
+    kind: "tool",
+    title: "工具调用 · web_search",
+    content: "query=\"任务关键词 最新动态\" → 命中 12 条，采纳 7 条（含来源与时间）",
+    detail:
+      '{"results":[{"title":"来源A：行业动态","snippet":"……"},{"title":"来源B：政策更新","snippet":"……"}]}',
+  });
+
+  if (info.state === "done") {
+    blocks.forEach((block) => {
+      messages.push({
+        kind: "artifact",
+        title: `产出 · ${block.name}`,
+        content: `（${block.name} 块内容摘要 · 演示）已按对方系统约定的结构生成，随本次执行写入时间线版本。`,
+      });
+    });
+  }
+
+  return { seatName, messages, blocks };
+}

frontend-web/src/pages/official-task-analysis/types.ts

@@ -0,0 +1,105 @@
+/**
+ * 外部任务多智能体定时分析 —— 前端先行数据模型（后端接口就绪后替换 store 层）。
+ *
+ * 对应设计定稿：
+ * - 外部任务（对方系统推送创建/删除）1:1 绑定「业务链 + 频率 + 块→席位映射」。
+ * - 执行记录两类：schedule 定时执行 / manual 手动执行（对方详情页总控问答）。
+ * - 时间线内容版本：每次定时执行产出各块内容（供对方详情接口拉取），管理页内
+ *   在席位执行过程抽屉里预览。
+ */
+
+import type { ScheduleConfig } from "@/components/workspace/scheduled-tasks/schedule-builder";
+
+export type TaRunTrigger = "schedule" | "manual";
+export type TaRunStatus = "pending" | "running" | "succeeded" | "failed";
+export type TaSeatState = "waiting" | "running" | "done" | "failed";
+
+/** 对方详情的分块内容与业务链席位的映射（存在绑定上，业务链保持纯粹）。 */
+export interface TaBlockMapping {
+  /** 块类型 key（对方系统定义）。 */
+  key: string;
+  /** 块展示名。 */
+  name: string;
+  /** 产出该块的席位（业务链 agent_id）。 */
+  seatAgentId: string;
+}
+
+export interface TaBinding {
+  chainId: string;
+  /** 业务链标题快照（展示用，链被删后仍可读）。 */
+  chainTitle: string;
+  cron: string;
+  cronText: string;
+  enabled: boolean;
+  blocks: TaBlockMapping[];
+  coordinatorPrompt?: string;
+  /** 问答沉淀的「执行方向备注」，注入后续每次定时执行（累积追加）。 */
+  directionNotes: string[];
+  /** 绑定时记住，作为下次创建任务的默认业务链返回给对方系统。 */
+  timezone: string;
+}
+
+export interface TaTask {
+  id: string;
+  /** 对方系统任务 id。 */
+  sourceTaskId: string;
+  name: string;
+  description: string;
+  binding: TaBinding | null;
+  createdAt: string;
+}
+
+/** 落库的运行记录；状态由 startedAt + durationMs 在读取时按当前时间推导。 */
+export interface TaRunRecord {
+  id: string;
+  taskId: string;
+  trigger: TaRunTrigger;
+  scheduledFor: string;
+  startedAt: string;
+  durationMs: number;
+  /** 该次执行的业务链编排快照（席位/阶段），链后续修改不影响历史回放。 */
+  stages: string[][];
+  /** 各阶段目标快照（与 stages 同下标），供流程图连线标签展示。 */
+  stageGoals?: (string | null)[] | null;
+  seatNames: Record<string, string>;
+  /** 在该阶段（下标）失败；null = 成功。 */
+  failStage: number | null;
+  /** 手动执行（总控问答）的摘要。 */
+  manualSummary?: string;
+}
+
+/** 读取时推导出的运行视图（每席位状态 + 耗时）。 */
+export interface TaSeatRunInfo {
+  state: TaSeatState;
+  durationMs: number;
+}
+
+export interface TaRunView {
+  id: string;
+  taskId: string;
+  trigger: TaRunTrigger;
+  scheduledFor: string;
+  startedAt: string;
+  finishedAt: string | null;
+  status: TaRunStatus;
+  stageIndex: number;
+  stageTotal: number;
+  seatInfos: Record<string, TaSeatRunInfo>;
+  error: string | null;
+  /** 各块产出状态（按绑定块映射推导；未绑定返回空）。 */
+  blockStates: Array<{ key: string; name: string; state: TaSeatState }>;
+  manualSummary?: string;
+  stages: string[][];
+  stageGoals?: (string | null)[] | null;
+  seatNames: Record<string, string>;
+}
+
+/** 绑定弹窗的提交载荷（内容块映射已移除；blocks 仅为兼容旧数据保留）。 */
+export interface TaBindingPayload {
+  chainId: string;
+  chainTitle: string;
+  schedule: ScheduleConfig;
+  enabled: boolean;
+  blocks?: TaBlockMapping[];
+  coordinatorPrompt?: string;
+}

frontend-web/src/position-roundtable/components/PositionIntentInput.tsx

@@ -26,6 +26,11 @@ interface PositionIntentInputProps extends Step1InputAreaProps {
   selectedModel?: string;
   onSelectedModelChange?: (model: string) => void;
   commonQuestions?: (string | AgentQuestion)[] | null;
+  /**
+   * 点击常用问题 chip 时直接把问题发送出去（8BF 第一步）。
+   * 缺省 false：仅填充输入框（岗位会商等既有页面的行为保持不变）。
+   */
+  sendCommonQuestionOnClick?: boolean;
   /** 仅保留上传、模型选择和发送的紧凑输入区（用于会商第 3 步）。 */
   hideUtilityActions?: boolean;
 }
@@ -68,6 +73,7 @@ export function PositionIntentInput({
   selectedModel,
   onSelectedModelChange,
   commonQuestions,
+  sendCommonQuestionOnClick = false,
   hideUtilityActions = false,
 }: PositionIntentInputProps) {
   // Files are owned by Step1Panel. This ref only opens the native picker and
@@ -86,7 +92,13 @@ export function PositionIntentInput({
     : (modelOptions[0]?.value ?? "");
 
   const pickCommonQuestion = (question: AgentQuestion) => {
-    onValueChange(question.prompt);
+    // 直接发送模式：以问题全文为覆盖文本调 onSend，绕过输入框 state 的
+    // 异步更新（同 tick 里 onSend 闭包读不到刚 setValue 的新值）。
+    if (sendCommonQuestionOnClick) {
+      onSend(question.prompt);
+    } else {
+      onValueChange(question.prompt);
+    }
     setShowMoreQuestions(false);
   };
 
@@ -160,6 +172,7 @@ export function PositionIntentInput({
                   type="button"
                   className="position-roundtable-intent-common-question"
                   title={question.prompt}
+                  disabled={sendCommonQuestionOnClick ? isSending : undefined}
                   onClick={() => pickCommonQuestion(question)}
                 >
                   <span aria-hidden="true" />
@@ -185,6 +198,7 @@ export function PositionIntentInput({
                           key={`${question.prompt}-${index}`}
                           type="button"
                           title={question.prompt}
+                          disabled={sendCommonQuestionOnClick ? isSending : undefined}
                           onClick={() => pickCommonQuestion(question)}
                         >
                           {question.title}
@@ -268,7 +282,7 @@ export function PositionIntentInput({
           <button
             id="send-message-btn"
             type="button"
-            onClick={isSending ? onStop : onSend}
+            onClick={() => (isSending ? onStop() : onSend())}
             disabled={isSending ? false : !canSend}
             aria-label={isSending ? "暂停生成" : "发送"}
             title={isSending ? "暂停生成" : "发送"}

frontend-web/src/report-collaboration/components/ReportCollaborationHeader.tsx

@@ -1,9 +1,12 @@
 "use client";
 
-import { ArrowLeft, FileText, MessageSquarePlus, Square } from "lucide-react";
+import { ArrowLeft, ClipboardList, FileText, MessageSquarePlus, Square } from "lucide-react";
 
 import type { PagePhase } from "@/report-collaboration/state/selectors";
 
+/** 计划模式覆盖的阶段：第一条消息起到工作流确认执行前。 */
+const PLAN_MODE_PHASES = new Set<PagePhase>(["clarifying", "planning", "proposal_ready"]);
+
 const PHASE_BADGES: Record<PagePhase, { label: string; color: string; bg: string }> = {
   empty: { label: "新会话", color: "var(--rc-text-secondary)", bg: "transparent" },
   clarifying: { label: "澄清需求", color: "var(--rc-waiting)", bg: "rgba(217, 119, 6, 0.1)" },
@@ -51,6 +54,16 @@ export function ReportCollaborationHeader({
         <span className="rc-header-title" title={title}>
           报告协作 / {title}
         </span>
+        {PLAN_MODE_PHASES.has(phase) ? (
+          <span
+            className="rc-chip"
+            title="规划师正在梳理需求并规划工作流，确认后开始执行"
+            style={{ background: "rgba(37, 99, 235, 0.1)", color: "var(--rc-accent)" }}
+          >
+            <ClipboardList className="size-3" aria-hidden />
+            计划模式
+          </span>
+        ) : null}
         <span className="rc-chip" style={{ background: badge.bg, color: badge.color }}>
           {badge.label}
         </span>

frontend-web/src/report-collaboration/components/ReportFlowCanvas.tsx

@@ -156,8 +156,23 @@ function CanvasInner({ plan, nodeStates, selectedNodeId, onSelectNode, emptyHint
   const edges: Edge[] = useMemo(() => {
     if (!plan) return [];
     return plan.edges.map((edge, index) => {
-      const targetState = nodeStates[edge.target];
+      // 边语义 = 数据正在流入 target 节点：target 执行中 → 流动动画；
-      const active = targetState?.status === "running" || targetState?.status === "validating";
+      // 失败红虚线 / 等待橙 / 已被替代降灰，其余静态。
+      const targetStatus = nodeStates[edge.target]?.status;
+      let className = "rc-edge";
+      let stroke = "var(--rc-flow-edge)";
+      if (targetStatus === "running" || targetStatus === "validating") {
+        className = "rc-edge-active";
+        stroke = "var(--rc-running)";
+      } else if (targetStatus === "failed") {
+        className = "rc-edge-failed";
+        stroke = "var(--rc-danger)";
+      } else if (targetStatus === "awaiting_input") {
+        className = "rc-edge-waiting";
+        stroke = "var(--rc-waiting)";
+      } else if (targetStatus === "superseded" || targetStatus === "interrupted" || targetStatus === "cancelled") {
+        className = "rc-edge-dimmed";
+      }
       return {
         id: `${edge.source}->${edge.target}-${index}`,
         source: edge.source,
@@ -165,11 +180,8 @@ function CanvasInner({ plan, nodeStates, selectedNodeId, onSelectNode, emptyHint
         label: edge.label ?? undefined,
         type: "default",
         animated: false,
-        className: active ? "rc-edge-active" : "rc-edge",
+        className,
-        style: {
+        style: { stroke, strokeWidth: 2 },
-          stroke: active ? "var(--rc-running)" : "var(--rc-flow-edge)",
-          strokeWidth: 2,
-        },
       };
     });
   }, [plan, nodeStates]);

frontend-web/src/report-collaboration/components/RequirementSummaryCard.tsx

@@ -0,0 +1,52 @@
+"use client";
+
+import { ClipboardCheck } from "lucide-react";
+
+import type { ReportRequirementSnapshot } from "@/report-collaboration/api/types";
+
+interface RequirementSummaryCardProps {
+  requirement: ReportRequirementSnapshot;
+  /** 需求仍可修改的提示（生成方案前）；方案已出后为 false。 */
+  editable?: boolean;
+}
+
+function Field({ label, value }: { label: string; value: string | null | undefined }) {
+  if (!value) return null;
+  return (
+    <div className="flex gap-2 text-[12px] leading-5">
+      <span className="w-16 shrink-0 text-[var(--rc-text-secondary)]">{label}</span>
+      <span className="min-w-0 flex-1 break-words">{value}</span>
+    </div>
+  );
+}
+
+/** 计划模式 · 需求摘要卡：结构化展示规划师确认的需求快照。 */
+export function RequirementSummaryCard({ requirement, editable = true }: RequirementSummaryCardProps) {
+  const angles = (requirement.required_angles ?? []).filter(Boolean).join("、");
+  return (
+    <div className="rc-card mx-auto w-full max-w-[560px] rounded-xl border border-[var(--rc-border)] bg-[var(--rc-bg-panel)] p-3.5">
+      <div className="mb-2 flex items-center gap-2">
+        <ClipboardCheck className="size-4" style={{ color: "var(--rc-accent)" }} aria-hidden />
+        <span className="text-[13px] font-semibold">需求摘要 · v{requirement.revision}</span>
+        <span className="rc-chip ml-auto" style={{ background: "rgba(37,99,235,.1)", color: "var(--rc-accent)" }}>
+          计划模式
+        </span>
+      </div>
+      <div className="flex flex-col gap-1">
+        <Field label="主题" value={requirement.topic} />
+        <Field label="目的" value={requirement.objective} />
+        <Field label="读者" value={requirement.audience} />
+        <Field label="角度" value={angles || null} />
+        <Field label="时间范围" value={requirement.time_range} />
+        <Field label="地域" value={requirement.geography} />
+        <Field label="篇幅" value={requirement.length_target} />
+        <Field label="语气" value={requirement.tone} />
+      </div>
+      {editable ? (
+        <p className="mt-2.5 border-t border-[var(--rc-border)] pt-2 text-[11px] leading-4 text-[var(--rc-text-secondary)]">
+          我正在基于这份需求为你规划工作流，稍候即可在对话中查看方案；想调整可直接在下方补充。
+        </p>
+      ) : null}
+    </div>
+  );
+}

frontend-web/src/report-collaboration/pages/ReportCollaborationPage.tsx

@@ -4,7 +4,7 @@ import { useCallback, useEffect, useMemo, useRef, useState } from "react";
 import { useLocation, useNavigate } from "react-router-dom";
 import type { Message } from "@langchain/langgraph-sdk";
 import { toast } from "sonner";
-import { FileText, Loader2, Sparkles } from "lucide-react";
+import { FileText, Sparkles } from "lucide-react";
 
 import { ArtifactsProvider } from "@/components/workspace/artifacts";
 import { ThreadContext } from "@/components/workspace/messages/context";
@@ -26,6 +26,7 @@ import {
 } from "@/report-collaboration/components/ReportPreviewDialog";
 import { ReportSandboxOverlay } from "@/report-collaboration/components/ReportSandboxOverlay";
 import { ReportSpeakerShell } from "@/report-collaboration/components/ReportSpeakerShell";
+import { RequirementSummaryCard } from "@/report-collaboration/components/RequirementSummaryCard";
 import { SessionHistoryPanel } from "@/report-collaboration/components/SessionHistoryPanel";
 import { useReportPanelLayout } from "@/report-collaboration/hooks/useReportPanelLayout";
 import { useReportCollaborationMessages } from "@/report-collaboration/hooks/useReportCollaborationMessages";
@@ -35,6 +36,7 @@ import {
   buildCommandConfirmCard,
   buildPlanGroupCard,
   buildReportAvailableCard,
+  buildRequirementReadyCard,
   readReportCollaborationCard,
 } from "@/report-collaboration/state/message-adapter";
 import { resolveReportSpeaker } from "@/report-collaboration/state/speaker";
@@ -247,11 +249,23 @@ export function ReportCollaborationSessionPage({ sessionId }: { sessionId: strin
     if (next) setFocusedAgentRunId(next);
   }, [agentTabs, focusPinned, focusedAgentRunId, tabIds]);
 
-  // 虚拟业务卡：方案组（已出方案且 run 未开始）/ 命令确认 / 报告入口。
+  // 虚拟业务卡：需求摘要（计划模式）/ 方案组（已出方案且 run 未开始）/ 命令确认 / 报告入口。
   // 采用方案后卡片保留，「开始协作」按钮随选定态启用；run 开始后消失。
   const runStarted = Boolean(state.run && state.run.plan_id);
   const extraCards = useMemo<Message[]>(() => {
     const cards: Message[] = [];
+    if (
+      state.requirement &&
+      state.plans.length === 0 &&
+      !runStarted &&
+      Object.values(state.messagesById).some(
+        (message) =>
+          message.role === "ai" &&
+          (message.metadata as Record<string, unknown> | null | undefined)?.ready_to_plan === true,
+      )
+    ) {
+      cards.push(buildRequirementReadyCard(state.requirement));
+    }
     if (
       state.plans.length > 0 &&
       !runStarted &&
@@ -341,39 +355,16 @@ export function ReportCollaborationSessionPage({ sessionId }: { sessionId: strin
     [sessionId, session],
   );
 
-  // 澄清完成（协调者已写入需求版本）但还没生成过方案：给出显式「发起规划」入口。
+  // 方案生成由后端在规划师判定 ready 后自动衔接（同一次请求内规划），
-  const requirementReady = useMemo(
+  // 前端无手动「生成候选方案」按钮；此处仅保留摘要卡的注入判定。
-    () =>
-      Object.values(state.messagesById).some(
-        (message) =>
-          message.role === "ai" &&
-          (message.metadata as Record<string, unknown> | null | undefined)?.ready_to_plan === true,
-      ),
-    [state.messagesById],
-  );
-  const showPlanRequestEntry =
-    state.session?.status === "clarifying" && state.plans.length === 0 && requirementReady;
-  const [planRequestBusy, setPlanRequestBusy] = useState(false);
-  const handleRequestPlans = useCallback(async () => {
-    setPlanRequestBusy(true);
-    try {
-      await getReportCollaborationDriver().requestPlans(
-        sessionId,
-        state.session?.requirement_revision ?? null,
-      );
-      await session.refreshSnapshot();
-      toast.success("已生成候选方案，可在对话中选择");
-    } catch (err) {
-      toast.error(err instanceof Error ? err.message : String(err));
-    } finally {
-      setPlanRequestBusy(false);
-    }
-  }, [sessionId, session, state.session?.requirement_revision]);
 
   const businessCardSlot = useCallback(
     (message: Message) => {
       const card = readReportCollaborationCard(message);
       if (!card) return null;
+      if (card.kind === "requirement_ready") {
+        return <RequirementSummaryCard requirement={card.requirement} editable={!runStarted} />;
+      }
       if (card.kind === "plan_group") {
         return (
           <ReportPlanCards
@@ -414,7 +405,7 @@ export function ReportCollaborationSessionPage({ sessionId }: { sessionId: strin
       }
       return null;
     },
-    [previewPlan, previewPlanId, session],
+    [previewPlan, previewPlanId, runStarted, session],
   );
 
   const speakerShell = useCallback(
@@ -565,43 +556,19 @@ export function ReportCollaborationSessionPage({ sessionId }: { sessionId: strin
                         : undefined
                     }
                     composer={
-                      <div className="flex flex-col">
+                      <ReportComposer
-                        {showPlanRequestEntry ? (
+                        phase={phase}
-                          <div className="mb-1.5 flex justify-center">
+                        busy={session.sendBusy}
-                            <button
+                        target={composerTarget}
-                              type="button"
+                        onClearTarget={() => setComposerTarget(null)}
-                              className="rc-btn"
+                        onSend={(text) =>
-                              style={{
+                          void session.send(text, {
-                                background: "var(--rc-accent)",
+                            nodeId: composerTarget?.nodeId ?? null,
-                                color: "#fff",
-                                borderColor: "var(--rc-accent)",
-                              }}
-                              disabled={planRequestBusy}
-                              onClick={() => void handleRequestPlans()}
-                            >
-                              {planRequestBusy ? (
-                                <Loader2 className="rc-spin size-3.5" aria-hidden />
-                              ) : (
-                                <Sparkles className="size-3.5" aria-hidden />
-                              )}
-                              {planRequestBusy ? "正在生成候选方案…" : "需求已就绪，生成候选方案"}
-                            </button>
-                          </div>
-                        ) : null}
-                        <ReportComposer
-                          phase={phase}
-                          busy={session.sendBusy}
-                          target={composerTarget}
-                          onClearTarget={() => setComposerTarget(null)}
-                          onSend={(text) =>
-                            void session.send(text, {
-                              nodeId: composerTarget?.nodeId ?? null,
                               agentRunId: composerTarget?.agentRunId ?? null,
                             })
                           }
                           focusSignal={focusSignal}
                         />
-                      </div>
                     }
                   />
                 }

frontend-web/src/report-collaboration/state/message-adapter.ts

@@ -21,6 +21,7 @@ import type {
   CollaborationCommand,
   CollaborationMessage,
   ReportPlanCandidate,
+  ReportRequirementSnapshot,
   ReportVersion,
 } from "@/report-collaboration/api/types";
 
@@ -113,6 +114,11 @@ export type ReportCollaborationCardPayload =
       selectedPlanId: string | null;
       requirementRevision: number;
     }
+  | {
+      kind: "requirement_ready";
+      requirement: ReportRequirementSnapshot;
+      revision: number;
+    }
   | {
       kind: "command_confirm";
       command: CollaborationCommand;
@@ -163,6 +169,17 @@ export function buildPlanGroupCard(
   });
 }
 
+/** 需求摘要卡（计划模式：需求已就绪、尚未生成方案）。 */
+export function buildRequirementReadyCard(
+  requirement: ReportRequirementSnapshot,
+): Message {
+  return buildCardMessage(`rc-card-requirement-r${requirement.revision}`, {
+    kind: "requirement_ready",
+    requirement,
+    revision: requirement.revision,
+  });
+}
+
 /** 命令确认卡（command.confirmation.required 之后）。 */
 export function buildCommandConfirmCard(command: CollaborationCommand): Message {
   return buildCardMessage(`rc-card-confirm-${command.command_id}`, {

frontend-web/src/report-collaboration/styles/report-collaboration.css

@@ -850,6 +850,21 @@
   animation: rc-edge-flow 0.5s linear infinite;
 }
 
+/* P1.4：失败红虚线 / 等待用户橙色 / 已替代·中断降灰 */
+.rc-edge-failed .react-flow__edge-path {
+  stroke: var(--rc-danger);
+  stroke-dasharray: 4 4;
+}
+
+.rc-edge-waiting .react-flow__edge-path {
+  stroke: var(--rc-waiting);
+}
+
+.rc-edge-dimmed .react-flow__edge-path {
+  stroke: var(--rc-flow-edge);
+  opacity: 0.45;
+}
+
 @keyframes rc-edge-flow {
   from {
     stroke-dashoffset: 10;

frontend-web/src/roundtable-planning/api/intent.ts

@@ -12,6 +12,9 @@ export type { ClarificationPayload };
 export const INTENT_AGENT_ID = "roundtable-intent";
 /** Dedicated Step 1 agent for the position-roundtable workspace. */
 export const POSITION_INTENT_AGENT_ID = "position-roundtable-intent";
+/** Dedicated Step 1 agent for the 8BF roundtable workspace (same prompt family
+ *  as `roundtable-intent`, independently tunable via the agent config UI). */
+export const EIGHTBF_INTENT_AGENT_ID = "eightbf-roundtable-intent";
 
 /**
  * Task-intent agent (Step 1) API. Mirror of multi-agent.ts but with a single

frontend-web/src/roundtable-planning/components/ChainFlowDiagram.tsx

@@ -39,11 +39,16 @@ export interface ChainFlowStage {
   goal?: string | null;
 }
 
+/** 节点执行状态（任务分析回放传入；缺省展示 Ready）。 */
+export type ChainFlowSeatState = "waiting" | "running" | "done" | "failed";
+
 export interface ChainFlowDiagramProps {
   stages: ChainFlowStage[];
   /** 点击席位节点卡片（总控节点不触发）。 */
   onNodeClick?: (agentId: string, name?: string) => void;
   className?: string;
+  /** 传入时节点卡片叠加执行状态徽标（任务分析的运行回放）；不影响原有展示。 */
+  seatStates?: Record<string, { state: ChainFlowSeatState; durationMs?: number }> | null;
 }
 
 // ── 布局常量（世界坐标，单位 px）─────────────────────────────────────────────
@@ -84,7 +89,7 @@ interface PlacedNode {
   rightX: number;
 }
 
-export function ChainFlowDiagram({ stages, onNodeClick, className }: ChainFlowDiagramProps) {
+export function ChainFlowDiagram({ stages, onNodeClick, className, seatStates }: ChainFlowDiagramProps) {
   // ── 列 / 节点 / 连线 / 标签布局 ──────────────────────────────────────────
   const layout = useMemo(() => {
     // 列：总控 + 各 Stage。
@@ -315,15 +320,43 @@ export function ChainFlowDiagram({ stages, onNodeClick, className }: ChainFlowDi
         </svg>
 
         {/* 节点卡片（参考「业务链条配置器」样式：左彩色描边 + 头像 + 名称 +
-            L-阶段徽标 + 同色角色行 + 底部 Ready 心跳条；图标仍用我方 getAvatar）。 */}
+            L-阶段徽标 + 同色角色行 + 底部 Ready 心跳条；图标仍用我方 getAvatar）。
+            seatStates 传入时（任务分析回放）叠加执行状态徽标与状态描边。 */}
         {layout.cols.flatMap((col, colIdx) =>
           col.nodes.map((n) => {
             const isCoordinator = n.id === COORDINATOR_ID;
             const nc = nodeColorOf(n.avatarType);
+            const seatState = !isCoordinator && seatStates ? seatStates[n.id] : undefined;
+            const state = seatState?.state;
+            const stateStyle = state
+              ? ({
+                  waiting: "text-muted-foreground",
+                  running: "text-amber-600 dark:text-amber-300",
+                  done: "text-emerald-600 dark:text-emerald-300",
+                  failed: "text-red-600 dark:text-red-400",
+                } as const)[state]
+              : null;
+            const stateLabel =
+              state === "waiting"
+                ? "等待"
+                : state === "running"
+                  ? "执行中"
+                  : state === "done"
+                    ? "完成"
+                    : state === "failed"
+                      ? "失败"
+                      : null;
+            const stateSeconds = seatState?.durationMs
+              ? Math.max(1, Math.round(seatState.durationMs / 1000))
+              : null;
             return (
               <div
                 key={n.id}
-                className={cn("absolute", !isCoordinator && "cursor-pointer")}
+                className={cn(
+                  "absolute",
+                  !isCoordinator && "cursor-pointer",
+                  state === "waiting" && "opacity-60",
+                )}
                 style={{ left: n.x, top: n.y, width: NODE_W, height: NODE_H }}
                 onPointerDown={(e) => e.stopPropagation()}
                 onClick={() => {
@@ -339,9 +372,15 @@ export function ChainFlowDiagram({ stages, onNodeClick, className }: ChainFlowDi
                     isCoordinator
                       ? "border-border bg-primary/[0.04]"
                       : "border-border hover:-translate-y-0.5 hover:shadow-[0_8px_22px_rgba(15,23,42,0.10)]",
+                    state === "running" && "border-amber-400/80",
+                    state === "failed" && "border-red-500/70",
                   )}
                   // 左侧 4px 彩色描边：内联上色，避免被 border-border 的四边色覆盖。
-                  style={{ borderLeftColor: nc.hex, borderLeftWidth: 4 }}
+                  style={{
+                    borderLeftColor: nc.hex,
+                    borderLeftWidth: 4,
+                    ...(state === "running" ? { boxShadow: "0 0 0 3px rgba(245,158,11,0.15)" } : {}),
+                  }}
                 >
                   {/* 上：头像 + 名称 + 阶段徽标 + 角色行 */}
                   <div className="flex items-start gap-2 overflow-hidden">
@@ -359,13 +398,34 @@ export function ChainFlowDiagram({ stages, onNodeClick, className }: ChainFlowDi
                     </div>
                   </div>
 
-                  {/* 下：身份 + Ready 心跳条 */}
+                  {/* 下：身份 + 状态行（回放模式显示执行状态，否则 Ready 心跳条） */}
                   <div className="flex items-center justify-between border-t border-border/60 pt-1 text-[8px] text-muted-foreground">
                     <span className="truncate">{isCoordinator ? "总控协调核心" : "智能体席位"}</span>
-                    <span className="ml-1.5 flex shrink-0 items-center gap-1 rounded bg-muted px-1 py-0.5 font-medium text-primary">
+                    {state && stateLabel ? (
-                      <span className="size-1 rounded-full bg-emerald-500" />
+                      <span
-                      Ready
+                        className={cn(
-                    </span>
+                          "ml-1.5 flex shrink-0 items-center gap-1 rounded px-1 py-0.5 font-medium",
+                          stateStyle,
+                        )}
+                      >
+                        <span
+                          className={cn(
+                            "size-1.5 rounded-full",
+                            state === "done" && "bg-emerald-500",
+                            state === "running" && "animate-pulse bg-amber-500",
+                            state === "failed" && "bg-red-500",
+                            state === "waiting" && "bg-muted-foreground/60",
+                          )}
+                        />
+                        {stateLabel}
+                        {stateSeconds != null && state !== "waiting" ? ` · ${stateSeconds}s` : ""}
+                      </span>
+                    ) : (
+                      <span className="ml-1.5 flex shrink-0 items-center gap-1 rounded bg-muted px-1 py-0.5 font-medium text-primary">
+                        <span className="size-1 rounded-full bg-emerald-500" />
+                        Ready
+                      </span>
+                    )}
                   </div>
                 </div>
               </div>
@@ -445,22 +505,33 @@ export function ChainFlowDiagram({ stages, onNodeClick, className }: ChainFlowDi
             style={{ width: mmW, height: mmH }}
             onClick={onMinimapClick}
           >
-            {/* 节点缩略块 */}
+            {/* 节点缩略块（回放模式按状态着色） */}
-            {layout.allNodes.map((n) => (
+            {layout.allNodes.map((n) => {
-              <div
+              const state = n.id === COORDINATOR_ID ? null : seatStates?.[n.id]?.state ?? null;
-                key={n.id}
+              return (
-                className={cn(
+                <div
-                  "absolute rounded-[1px]",
+                  key={n.id}
-                  n.id === COORDINATOR_ID ? "bg-primary/70" : "bg-foreground/35",
+                  className={cn(
-                )}
+                    "absolute rounded-[1px]",
-                style={{
+                    n.id === COORDINATOR_ID
-                  left: n.x * mmScale,
+                      ? "bg-primary/70"
-                  top: n.y * mmScale,
+                      : state === "done"
-                  width: NODE_W * mmScale,
+                        ? "bg-emerald-500/70"
-                  height: NODE_H * mmScale,
+                        : state === "running"
-                }}
+                          ? "bg-amber-500/80"
-              />
+                          : state === "failed"
-            ))}
+                            ? "bg-red-500/80"
+                            : "bg-foreground/35",
+                  )}
+                  style={{
+                    left: n.x * mmScale,
+                    top: n.y * mmScale,
+                    width: NODE_W * mmScale,
+                    height: NODE_H * mmScale,
+                  }}
+                />
+              );
+            })}
             {/* 当前视口框 */}
             <div
               className="absolute border-2 border-primary bg-primary/10"

frontend-web/src/roundtable-planning/components/Step1Panel.tsx

@@ -81,7 +81,8 @@ export interface Step1InputAreaProps {
   isSending: boolean;
   canSend: boolean;
   pendingClarification: boolean;
-  onSend: () => void;
+  /** 发送当前输入框内容；传 `overrideText` 时直接发送该文本（常用问题一键发送）。 */
+  onSend: (overrideText?: string) => void;
   onStop: () => void;
   onFollowUp: () => void;
   onSkip: () => void;
@@ -236,9 +237,12 @@ export function Step1Panel(props: Step1PanelProps) {
     setAttachedFiles((prev) => prev.filter((_, i) => i !== idx));
   };
 
-  const handleSend = () => {
+  const handleSend = (overrideText?: string) => {
-    if (!inputText.trim() && attachedFiles.length === 0) return;
+    // overrideText：常用问题 chip 一键发送——绕过异步的输入框 state，
-    onSendMessage(inputText, attachedFiles.length > 0 ? attachedFiles : undefined);
+    // 直接以问题全文作为本回合消息（仍共享附件与发送后的清空逻辑）。
+    const textToSend = overrideText ?? inputText;
+    if (!textToSend.trim() && attachedFiles.length === 0) return;
+    onSendMessage(textToSend, attachedFiles.length > 0 ? attachedFiles : undefined);
     handleInputTextChange("");
     setAttachedFiles([]);
   };
@@ -411,7 +415,7 @@ export function Step1Panel(props: Step1PanelProps) {
           <div className="flex items-center gap-2">
             <button
               id="send-message-btn"
-              onClick={handleSend}
+              onClick={() => handleSend()}
               disabled={isIntentSending || !intentThreadId}
               className="bg-primary text-white hover:bg-[var(--brand-900)] disabled:bg-slate-300 disabled:cursor-not-allowed text-xs font-bold px-4 py-2.5 rounded-xl flex items-center gap-1.5 transition-all shadow-[0_2px_8px_rgba(245,34,45,0.15)] active:scale-95"
             >

frontend-web/src/roundtable-planning/hooks/useStep2Orchestration.ts

@@ -422,6 +422,12 @@ export interface UseStep2OrchestrationOptions {
   seatSkillDirectives?: Record<string, boolean>;
   /** Per-seat human-validation checkpoints inherited from the selected chain. */
   humanValidations?: Record<string, boolean>;
+  /**
+   * 全局「人工干预」开关（8BF 页头）：开启后第二步每走一步（链条每席位 / DAG 每 stage
+   * 完成）就暂停等待用户输入或恢复；关闭则全自动推进，与既有行为一致。与 per-seat
+   * `humanValidations` 取 OR——任一生效即在最近一步边界暂停。
+   */
+  manualIntervention?: boolean;
   /**
    * 每席位「推理深度覆盖」映射 {agentId: SeatMode | null}（沿用所选业务链条里每个席位的配置）。
    * 派活该 special 席位时按 agentId 查表，**席位覆盖优先、否则跟随链条级 `selectedMode`**：
@@ -593,6 +599,7 @@ export function useStep2Orchestration(
     selectedMode,
     seatSkillDirectives = {},
     humanValidations = {},
+    manualIntervention = false,
     seatModes = {},
     gatherFirst = false,
     orchestrationMode,
@@ -746,6 +753,9 @@ export function useStep2Orchestration(
   // 每席位技能识别强化映射（所选业务链条各席位开关）同样打到 ref，席位调用按 agentId 读最新值。
   const seatSkillDirectivesRef = useRef<Record<string, boolean>>(seatSkillDirectives);
   const humanValidationsRef = useRef<Record<string, boolean>>(humanValidations);
+  // 全局「人工干预」开关（8BF 页头）：编排循环在每个一步边界读最新值，
+  // 中途切换只影响下一次边界，不打断在途 run。
+  const manualInterventionRef = useRef<boolean>(manualIntervention);
   // 每席位推理深度覆盖映射（所选业务链条各席位 seatMode）打到 ref，席位调用按 agentId 读最新值；
   // 覆盖优先、否则跟随 selectedModeRef。
   const seatModesRef = useRef<Record<string, SeatMode | null>>(seatModes);
@@ -800,6 +810,9 @@ export function useStep2Orchestration(
   useEffect(() => {
     humanValidationsRef.current = humanValidations;
   }, [humanValidations]);
+  useEffect(() => {
+    manualInterventionRef.current = manualIntervention;
+  }, [manualIntervention]);
   useEffect(() => {
     seatModesRef.current = seatModes;
   }, [seatModes]);
@@ -1681,12 +1694,16 @@ export function useStep2Orchestration(
   );
 
   const pauseForHumanValidation = useCallback(
-    (continuation: HumanValidationContinuation, seatNames: string[]) => {
+    (
+      continuation: HumanValidationContinuation,
+      seatNames: string[],
+      source = "业务链条配置",
+    ) => {
       humanValidationContinuationRef.current = continuation;
       appendDialogue({
         sender: "系统",
         avatarType: "blue",
-        content: `已完成 ${seatNames.join("、")} 的输出，按业务链条配置暂停，等待人工校验。可直接点击“恢复会商”继续，或在下方输入修改意见后提交。`,
+        content: `已完成 ${seatNames.join("、")} 的输出，按${source}暂停，等待人工校验。可直接点击“恢复会商”继续，或在下方输入修改意见后提交。`,
         confidence: "等待人工校验",
       });
       setIsRoundTableRunning(false);
@@ -2793,18 +2810,27 @@ export function useStep2Orchestration(
             }
             if (!chainInvalidNames.includes(seatName)) chainInvalidNames.push(seatName);
             finishedAllSeats = false;
-            if (humanValidationsRef.current[seat.agent_id]) {
+            if (manualInterventionRef.current || humanValidationsRef.current[seat.agent_id]) {
-              pauseForHumanValidation({ mode: "chain", nextSeatIndex: i }, [seatName]);
+              pauseForHumanValidation(
+                { mode: "chain", nextSeatIndex: i },
+                [seatName],
+                manualInterventionRef.current && !humanValidationsRef.current[seat.agent_id]
+                  ? "人工干预开关"
+                  : undefined,
+              );
               break;
             }
             continue;
           }
 
           setConsensusPercentage(Math.min(95, Math.round(((i + 1) / total) * 90)));
-          if (humanValidationsRef.current[seat.agent_id]) {
+          if (manualInterventionRef.current || humanValidationsRef.current[seat.agent_id]) {
             pauseForHumanValidation(
               { mode: "chain", nextSeatIndex: i + 1 },
               [seatName],
+              manualInterventionRef.current && !humanValidationsRef.current[seat.agent_id]
+                ? "人工干预开关"
+                : undefined,
             );
             finishedAllSeats = false;
             break;
@@ -3237,8 +3263,11 @@ export function useStep2Orchestration(
           setConsensusPercentage(
             Math.min(95, Math.round(((s + 1) / plan.stages.length) * 90)),
           );
+          const manualInterventionOn = manualInterventionRef.current;
           const validationSeatNames = stageSeats
-            .filter(({ agentId }) => humanValidationsRef.current[agentId])
+            .filter(
+              ({ agentId }) => manualInterventionOn || humanValidationsRef.current[agentId],
+            )
             .map(({ name }) => name);
           if (validationSeatNames.length > 0) {
             pauseForHumanValidation(
@@ -3248,6 +3277,7 @@ export function useStep2Orchestration(
                 priorDeliveries: [...priorDeliveries],
               },
               validationSeatNames,
+              manualInterventionOn ? "人工干预开关" : undefined,
             );
             finishedAllStages = false;
             break;

frontend-web/src/roundtable-planning/pages/RoundtablePlanningPage.tsx

@@ -75,6 +75,7 @@ import {
   DropdownMenuTrigger,
 } from "@/components/ui/dropdown-menu";
 import { Tooltip, TooltipContent, TooltipTrigger } from "@/components/ui/tooltip";
+import { Switch } from "@/components/ui/switch";
 import { desensitize } from "@/lib/desensitize";
 import { cn } from "@/lib/utils";
 import { useDesensitizeSubtree } from "@/roundtable-planning/lib/use-desensitize-subtree";
@@ -324,6 +325,15 @@ export default function RoundtablePlanningPage({
   entryRailTaskCardSlot,
   /** 独立业务页传给意图智能体的隐藏任务上下文。 */
   taskContext,
+  /** 第一步绑定的内置意图智能体 id（如 8BF 的 eightbf-roundtable-intent）。
+   *  缺省沿用通用会商的 roundtable-intent，原多智能体会商入口行为不变。 */
+  intentAgentId,
+  /**
+   * 8BF 页头「人工干预」开关（受控）：开启后第二步每走一步暂停等待用户输入，
+   * 关闭自动推进。提供 onStep2InterventionChange 时才渲染开关，通用会商页不受影响。
+   */
+  step2InterventionEnabled,
+  onStep2InterventionChange,
   /** 独立业务页的任务作用域，仅用于左栏记录筛选，不改变原深链逻辑。 */
   entryRailTaskId,
   /** 独立业务页的根节点样式类，避免修改原会商样式。 */
@@ -367,6 +377,11 @@ export default function RoundtablePlanningPage({
   entryRailBusinessChainSlot?: ReactNode;
   entryRailTaskCardSlot?: ReactNode;
   taskContext?: Record<string, unknown> | null;
+  /** 第一步绑定的内置意图智能体 id；缺省走通用 roundtable-intent。 */
+  intentAgentId?: string;
+  /** 第二步「人工干预」开关（8BF 页头，新建任务按钮左侧）。 */
+  step2InterventionEnabled?: boolean;
+  onStep2InterventionChange?: (enabled: boolean) => void;
   entryRailTaskId?: string | null;
   rootClassName?: string;
   renderStep1?: (props: Step1PanelProps & {
@@ -755,6 +770,7 @@ export default function RoundtablePlanningPage({
     selectedMode: seatMode,
     seatSkillDirectives,
     humanValidations,
+    manualIntervention: step2InterventionEnabled ?? false,
     seatModes,
     gatherFirst,
     orchestrationMode,
@@ -771,6 +787,7 @@ export default function RoundtablePlanningPage({
     selectedModel,
     triggerToast,
     taskContext,
+    intentAgentId,
   });
 
   const taskArtifactSubmission = useTaskArtifactSubmission(
@@ -2381,6 +2398,23 @@ export default function RoundtablePlanningPage({
             </button>
           )}
 
+          {/* 8BF：第二步「人工干预」开关。开启后第二步每走一步暂停，等待用户输入 /
+              恢复后继续；关闭则自动推进（与原行为一致）。仅独立业务页传入受控回调时渲染。 */}
+          {onStep2InterventionChange && (
+            <label
+              id="step2-intervention-toggle"
+              title="开启后：第二步每完成一步暂停，可在下方输入修改意见或点击“恢复会商”继续；关闭后：自动完成整个研讨"
+              className="bg-white hover:bg-slate-50 border border-[#e2e8f0] text-[#334155] text-xs font-bold px-3 py-2 rounded-xl flex items-center gap-2 transition cursor-pointer select-none dark:border-border dark:bg-secondary dark:text-foreground dark:hover:bg-accent"
+            >
+              <Switch
+                checked={step2InterventionEnabled ?? false}
+                onCheckedChange={onStep2InterventionChange}
+                className="scale-90"
+              />
+              人工干预
+            </label>
+          )}
+
           {/* 深链接（taskId）模式：任务固定，把「新建任务」换成「重新分析」——清空会话、对同一任务重做意图分析。 */}
           {prefillRef.current?.taskId ? (
             <button

frontend-web/src/styles/agents-orbit-dark.css

@@ -0,0 +1,168 @@
+/* ============================================================
+   智能体星轨页 — 深色主题（星海蓝）
+   只放颜色，结构与动画在 agents-orbit.css
+   ============================================================ */
+
+.agents-orbit-page.ao-dark {
+  color: #d9e8ff;
+  background:
+    radial-gradient(1100px 640px at 22% 40%, rgba(56, 130, 246, 0.18), transparent 62%),
+    radial-gradient(880px 560px at 80% 26%, rgba(139, 92, 246, 0.14), transparent 60%),
+    linear-gradient(158deg, #050b1e 0%, #081429 46%, #03060f 100%);
+}
+
+/* 星点 */
+.agents-orbit-page.ao-dark::before {
+  background-image:
+    radial-gradient(1.6px 1.6px at 11% 16%, rgba(255, 255, 255, 0.55), transparent 55%),
+    radial-gradient(1.2px 1.2px at 27% 74%, rgba(255, 255, 255, 0.35), transparent 55%),
+    radial-gradient(1.4px 1.4px at 42% 26%, rgba(255, 255, 255, 0.4), transparent 55%),
+    radial-gradient(1px 1px at 58% 82%, rgba(255, 255, 255, 0.3), transparent 55%),
+    radial-gradient(1.6px 1.6px at 66% 12%, rgba(255, 255, 255, 0.5), transparent 55%),
+    radial-gradient(1.2px 1.2px at 79% 58%, rgba(255, 255, 255, 0.35), transparent 55%),
+    radial-gradient(1px 1px at 88% 34%, rgba(255, 255, 255, 0.3), transparent 55%),
+    radial-gradient(1.4px 1.4px at 18% 88%, rgba(255, 255, 255, 0.35), transparent 55%),
+    radial-gradient(1px 1px at 35% 48%, rgba(255, 255, 255, 0.25), transparent 55%),
+    radial-gradient(1.2px 1.2px at 92% 78%, rgba(255, 255, 255, 0.3), transparent 55%);
+}
+
+/* 顶部标题 */
+.agents-orbit-page.ao-dark .ao-header-title {
+  color: #cfe4ff;
+  text-shadow: 0 0 14px rgba(90, 160, 255, 0.5);
+}
+
+.agents-orbit-page.ao-dark .ao-header-link {
+  border-color: rgba(110, 170, 255, 0.35);
+  background: rgba(12, 28, 58, 0.7);
+  color: #9fc4f5;
+}
+
+.agents-orbit-page.ao-dark .ao-header-link:hover {
+  color: #ffffff;
+  border-color: rgba(140, 205, 255, 0.8);
+  box-shadow: 0 0 14px rgba(90, 170, 255, 0.4);
+}
+
+/* 中心圆盘 */
+.agents-orbit-page.ao-dark .ao-core {
+  background: radial-gradient(
+    circle at 34% 30%,
+    rgba(96, 165, 250, 0.95),
+    rgba(37, 99, 235, 0.55) 45%,
+    rgba(8, 20, 45, 0.92) 100%
+  );
+  border-color: rgba(125, 190, 255, 0.7);
+  box-shadow:
+    0 0 46px rgba(64, 140, 255, 0.5),
+    inset 0 0 34px rgba(120, 190, 255, 0.35);
+}
+
+.agents-orbit-page.ao-dark .ao-core-img {
+  color: #eaf4ff;
+  filter: drop-shadow(0 0 14px rgba(120, 190, 255, 0.85));
+}
+
+/* 圆环族 */
+.agents-orbit-page.ao-dark .ao-ring--dash {
+  border-color: rgba(110, 170, 255, 0.55);
+}
+
+.agents-orbit-page.ao-dark .ao-ring--arc {
+  border-top-color: rgba(88, 200, 255, 0.9);
+  border-right-color: rgba(88, 200, 255, 0.32);
+  filter: drop-shadow(0 0 6px rgba(88, 200, 255, 0.55));
+}
+
+.agents-orbit-page.ao-dark .ao-ring--ticks {
+  border-color: rgba(110, 170, 255, 0.28);
+  background-image: repeating-conic-gradient(
+    from 0deg,
+    rgba(120, 180, 255, 0.5) 0deg 0.8deg,
+    transparent 0.8deg 6deg
+  );
+}
+
+/* 卫星 */
+.agents-orbit-page.ao-dark .ao-sat {
+  background: radial-gradient(circle at 35% 30%, #ffffff, #7fd0ff 55%, #1f7dff);
+  box-shadow:
+    0 0 12px 3px rgba(110, 200, 255, 0.85),
+    -14px 0 22px rgba(110, 200, 255, 0.35);
+}
+
+/* 类型标签药丸 */
+.agents-orbit-page.ao-dark .ao-type-chip {
+  color: #b9d7ff;
+  background: rgba(12, 28, 58, 0.82);
+  border-color: rgba(110, 170, 255, 0.35);
+  box-shadow: 0 4px 16px rgba(4, 12, 32, 0.55);
+}
+
+.agents-orbit-page.ao-dark .ao-type-anchor:hover .ao-type-chip {
+  color: #ffffff;
+  border-color: rgba(140, 205, 255, 0.8);
+  box-shadow: 0 0 18px rgba(90, 170, 255, 0.45);
+}
+
+.agents-orbit-page.ao-dark .ao-type-anchor.is-active .ao-type-chip {
+  color: #06122b;
+  background: linear-gradient(135deg, #66c4ff, #2f7dff);
+  border-color: rgba(160, 220, 255, 0.9);
+  box-shadow: 0 0 26px rgba(80, 170, 255, 0.65);
+}
+
+/* 右侧面板 */
+.agents-orbit-page.ao-dark .ao-panel {
+  background: rgba(10, 22, 48, 0.72);
+  border-color: rgba(110, 170, 255, 0.22);
+  box-shadow: 0 10px 40px rgba(2, 8, 24, 0.5);
+}
+
+.agents-orbit-page.ao-dark .ao-panel-head {
+  border-bottom-color: rgba(110, 170, 255, 0.16);
+}
+
+.agents-orbit-page.ao-dark .ao-panel-title {
+  color: #e8f2ff;
+}
+
+.agents-orbit-page.ao-dark .ao-panel-sub {
+  color: #8fb7e8;
+}
+
+.agents-orbit-page.ao-dark .ao-agents::-webkit-scrollbar-thumb {
+  background: rgba(110, 170, 255, 0.3);
+}
+
+.agents-orbit-page.ao-dark .ao-agent {
+  border-color: rgba(110, 170, 255, 0.18);
+  background: rgba(20, 42, 86, 0.5);
+}
+
+.agents-orbit-page.ao-dark .ao-agent:hover {
+  border-color: rgba(130, 200, 255, 0.6);
+  background: rgba(30, 58, 112, 0.65);
+  box-shadow: 0 8px 24px rgba(20, 80, 200, 0.35);
+}
+
+.agents-orbit-page.ao-dark .ao-agent-icon {
+  background: linear-gradient(135deg, #2f7dff, #7a5cff);
+  box-shadow:
+    inset 0 0 0 1px rgba(255, 255, 255, 0.25),
+    0 4px 14px rgba(47, 125, 255, 0.4);
+}
+
+.agents-orbit-page.ao-dark .ao-agent-name {
+  color: #d7e8ff;
+}
+
+/* 加载 / 空态 */
+.agents-orbit-page.ao-dark .ao-state {
+  color: #8fb7e8;
+}
+
+.agents-orbit-page.ao-dark .ao-spinner {
+  border-color: rgba(110, 170, 255, 0.25);
+  border-top-color: #6fc2ff;
+}

frontend-web/src/styles/agents-orbit-light.css

@@ -0,0 +1,158 @@
+/* ============================================================
+   智能体星轨页 — 浅色主题（云际蓝）
+   只放颜色，结构与动画在 agents-orbit.css
+   ============================================================ */
+
+.agents-orbit-page.ao-light {
+  color: #1f3a63;
+  background:
+    radial-gradient(1100px 640px at 22% 40%, rgba(47, 125, 255, 0.1), transparent 62%),
+    radial-gradient(880px 560px at 80% 26%, rgba(122, 92, 246, 0.08), transparent 60%),
+    linear-gradient(158deg, #f4f8ff 0%, #e7f0ff 46%, #eef4fd 100%);
+}
+
+/* 浅色下不渲染星点 */
+.agents-orbit-page.ao-light::before {
+  display: none;
+}
+
+/* 顶部标题 */
+.agents-orbit-page.ao-light .ao-header-title {
+  color: #164a96;
+  text-shadow: 0 0 14px rgba(120, 170, 255, 0.35);
+}
+
+.agents-orbit-page.ao-light .ao-header-link {
+  border-color: rgba(47, 125, 255, 0.4);
+  background: rgba(255, 255, 255, 0.85);
+  color: #2a66c2;
+}
+
+.agents-orbit-page.ao-light .ao-header-link:hover {
+  color: #0f3f96;
+  border-color: rgba(47, 125, 255, 0.85);
+  box-shadow: 0 4px 14px rgba(47, 125, 255, 0.22);
+}
+
+/* 中心圆盘 */
+.agents-orbit-page.ao-light .ao-core {
+  background: radial-gradient(
+    circle at 34% 30%,
+    #ffffff,
+    #dbe9ff 45%,
+    #b9d4ff 100%
+  );
+  border-color: rgba(90, 150, 240, 0.85);
+  box-shadow:
+    0 10px 34px rgba(47, 125, 255, 0.28),
+    inset 0 0 30px rgba(160, 200, 255, 0.55);
+}
+
+.agents-orbit-page.ao-light .ao-core-img {
+  color: #1f5fd6;
+  filter: drop-shadow(0 4px 10px rgba(47, 125, 255, 0.35));
+}
+
+/* 圆环族 */
+.agents-orbit-page.ao-light .ao-ring--dash {
+  border-color: rgba(47, 125, 255, 0.45);
+}
+
+.agents-orbit-page.ao-light .ao-ring--arc {
+  border-top-color: rgba(37, 110, 245, 0.85);
+  border-right-color: rgba(37, 110, 245, 0.28);
+  filter: drop-shadow(0 2px 5px rgba(47, 125, 255, 0.3));
+}
+
+.agents-orbit-page.ao-light .ao-ring--ticks {
+  border-color: rgba(47, 125, 255, 0.3);
+  background-image: repeating-conic-gradient(
+    from 0deg,
+    rgba(47, 105, 220, 0.45) 0deg 0.8deg,
+    transparent 0.8deg 6deg
+  );
+}
+
+/* 卫星 */
+.agents-orbit-page.ao-light .ao-sat {
+  background: radial-gradient(circle at 35% 30%, #ffffff, #7fb4ff 55%, #2f7dff);
+  box-shadow:
+    0 0 10px 2px rgba(47, 125, 255, 0.55),
+    -14px 0 20px rgba(47, 125, 255, 0.25);
+}
+
+/* 类型标签药丸 */
+.agents-orbit-page.ao-light .ao-type-chip {
+  color: #1e4fa8;
+  background: rgba(255, 255, 255, 0.92);
+  border-color: rgba(47, 125, 255, 0.4);
+  box-shadow: 0 4px 14px rgba(30, 80, 180, 0.14);
+}
+
+.agents-orbit-page.ao-light .ao-type-anchor:hover .ao-type-chip {
+  color: #0f3f96;
+  border-color: rgba(47, 125, 255, 0.85);
+  box-shadow: 0 6px 18px rgba(47, 125, 255, 0.28);
+}
+
+.agents-orbit-page.ao-light .ao-type-anchor.is-active .ao-type-chip {
+  color: #ffffff;
+  background: linear-gradient(135deg, #4d9aff, #2f7dff);
+  border-color: rgba(47, 125, 255, 0.95);
+  box-shadow: 0 8px 24px rgba(47, 125, 255, 0.45);
+}
+
+/* 右侧面板 */
+.agents-orbit-page.ao-light .ao-panel {
+  background: rgba(255, 255, 255, 0.82);
+  border-color: rgba(47, 125, 255, 0.25);
+  box-shadow: 0 10px 32px rgba(30, 70, 160, 0.12);
+}
+
+.agents-orbit-page.ao-light .ao-panel-head {
+  border-bottom-color: rgba(47, 125, 255, 0.18);
+}
+
+.agents-orbit-page.ao-light .ao-panel-title {
+  color: #0f3a78;
+}
+
+.agents-orbit-page.ao-light .ao-panel-sub {
+  color: #5b7ea8;
+}
+
+.agents-orbit-page.ao-light .ao-agents::-webkit-scrollbar-thumb {
+  background: rgba(47, 125, 255, 0.3);
+}
+
+.agents-orbit-page.ao-light .ao-agent {
+  border-color: rgba(47, 125, 255, 0.22);
+  background: rgba(255, 255, 255, 0.85);
+}
+
+.agents-orbit-page.ao-light .ao-agent:hover {
+  border-color: rgba(47, 125, 255, 0.6);
+  background: #f0f7ff;
+  box-shadow: 0 8px 22px rgba(47, 125, 255, 0.22);
+}
+
+.agents-orbit-page.ao-light .ao-agent-icon {
+  background: linear-gradient(135deg, #2f7dff, #7a5cff);
+  box-shadow:
+    inset 0 0 0 1px rgba(255, 255, 255, 0.35),
+    0 4px 12px rgba(47, 125, 255, 0.35);
+}
+
+.agents-orbit-page.ao-light .ao-agent-name {
+  color: #1f3a63;
+}
+
+/* 加载 / 空态 */
+.agents-orbit-page.ao-light .ao-state {
+  color: #5b7ea8;
+}
+
+.agents-orbit-page.ao-light .ao-spinner {
+  border-color: rgba(47, 125, 255, 0.25);
+  border-top-color: #2f7dff;
+}

frontend-web/src/styles/agents-orbit.css

@@ -0,0 +1,432 @@
+/* ============================================================
+   智能体星轨页 — 基础结构 + 动画（不含任何颜色）
+   颜色主题在 agents-orbit-dark.css / agents-orbit-light.css 两个文件里
+   命名统一 ao- 前缀
+   ============================================================ */
+
+/* ---------- 页面骨架：左舞台 / 右面板 1:1 ---------- */
+
+.agents-orbit-page {
+  position: relative;
+  display: flex;
+  height: 100%;
+  width: 100%;
+  overflow: hidden;
+}
+
+/* 星点背景层（内容在主题文件里着色；浅色主题直接隐藏） */
+.agents-orbit-page::before {
+  content: "";
+  position: absolute;
+  inset: 0;
+  pointer-events: none;
+  animation: ao-star-blink 7s ease-in-out infinite;
+}
+
+@keyframes ao-star-blink {
+  0%,
+  100% {
+    opacity: 0.65;
+  }
+  50% {
+    opacity: 1;
+  }
+}
+
+/* 顶部小标题 + 返回经典视图入口 */
+.ao-header {
+  position: absolute;
+  top: 18px;
+  left: 22px;
+  z-index: 6;
+  display: flex;
+  align-items: center;
+  gap: 14px;
+}
+
+.ao-header-title {
+  font-size: 15px;
+  font-weight: 600;
+  letter-spacing: 2px;
+}
+
+.ao-header-link {
+  border: 1px solid;
+  background: none;
+  font-size: 12px;
+  padding: 4px 12px;
+  border-radius: 999px;
+  cursor: pointer;
+  transition: color 0.25s, border-color 0.25s, box-shadow 0.25s;
+}
+
+/* ---------- 左侧星轨舞台 ---------- */
+
+.ao-stage {
+  position: relative;
+  flex: 1 1 0;
+  min-width: 0;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+}
+
+.ao-root {
+  position: relative;
+  width: min(56vh, 520px);
+  max-width: calc(100% - 48px);
+  aspect-ratio: 1 / 1;
+}
+
+/* 中心圆盘（核心） */
+.ao-core {
+  position: absolute;
+  left: 50%;
+  top: 50%;
+  width: 32%;
+  aspect-ratio: 1 / 1;
+  transform: translate(-50%, -50%);
+  border-radius: 50%;
+  border: 2px solid;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  z-index: 3;
+}
+
+/* 中心图片：上下摇晃（频率适中，约 3.2s 一个来回） */
+.ao-core-img {
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  animation: ao-float 3.2s ease-in-out infinite;
+}
+
+@keyframes ao-float {
+  0%,
+  100% {
+    transform: translateY(-7px);
+  }
+  50% {
+    transform: translateY(9px);
+  }
+}
+
+/* ---------- 圆环族 ----------
+   由内到外：虚线环(54%) 顺时针 / 弧光环(70%) 逆时针 / 刻度环(97%) 缓慢顺时针 */
+
+.ao-ring {
+  position: absolute;
+  left: 50%;
+  top: 50%;
+  transform: translate(-50%, -50%);
+  border-radius: 50%;
+  pointer-events: none;
+}
+
+.ao-ring--dash {
+  width: 54%;
+  aspect-ratio: 1 / 1;
+  border: 2px dashed;
+  animation: ao-spin-cw 18s linear infinite;
+}
+
+.ao-ring--arc {
+  width: 70%;
+  aspect-ratio: 1 / 1;
+  border: 3px solid transparent;
+  animation: ao-spin-ccw 12s linear infinite;
+}
+
+/* 最外层刻度圆圈：细边框 + 环形遮罩刻度（密码转盘的刻度感），
+   缓慢自转，转盘属性刻度环 */
+.ao-ring--ticks {
+  width: 97%;
+  aspect-ratio: 1 / 1;
+  border: 1px solid;
+  -webkit-mask: radial-gradient(
+    farthest-side,
+    transparent calc(100% - 5px),
+    #000 calc(100% - 4px)
+  );
+  mask: radial-gradient(
+    farthest-side,
+    transparent calc(100% - 5px),
+    #000 calc(100% - 4px)
+  );
+  animation: ao-spin-cw 90s linear infinite;
+}
+
+@keyframes ao-spin-cw {
+  to {
+    transform: translate(-50%, -50%) rotate(360deg);
+  }
+}
+
+@keyframes ao-spin-ccw {
+  to {
+    transform: translate(-50%, -50%) rotate(-360deg);
+  }
+}
+
+/* 卫星：一个自转的轨道容器，卫星固定在轨道顶端；
+   轨道直径与最外层刻度圆圈(97%)一致，卫星贴着最外圈环绕 */
+.ao-sat-orbit {
+  position: absolute;
+  left: 50%;
+  top: 50%;
+  width: 97%;
+  aspect-ratio: 1 / 1;
+  transform: translate(-50%, -50%);
+  pointer-events: none;
+  z-index: 2;
+  animation: ao-orbit-cw 9s linear infinite;
+}
+
+.ao-sat {
+  position: absolute;
+  top: -7px;
+  left: 50%;
+  width: 14px;
+  height: 14px;
+  margin-left: -7px;
+  border-radius: 50%;
+}
+
+@keyframes ao-orbit-cw {
+  to {
+    transform: translate(-50%, -50%) rotate(360deg);
+  }
+}
+
+/* ---------- 类型标签（密码转盘） ---------- */
+
+/* 转盘裁剪容器：垂直中线以左（左半圆）是不可视区域，
+   绕背补位的标签滑出边界即被裁掉、从另一端滑入才重新出现。
+   上/右/下方向用负 inset 放宽裁剪框，不切到溢出 root 的标签。 */
+.ao-types {
+  position: absolute;
+  inset: 0;
+  z-index: 4;
+  clip-path: inset(-80px -80px -80px 50%);
+}
+
+/* 旋转臂：铺满整个 root、以圆心为轴旋转；标签挂在 93%/50% 处（刻度环内侧）。
+   换角度时 CSS 过渡让标签沿圆弧运动，文字由锚点反向旋转保持水平。
+   超界补位的标签角度会 ±360 走“背面长路”，全程经过被裁剪的左半圆。 */
+.ao-type-arm {
+  position: absolute;
+  inset: 0;
+  pointer-events: none;
+  transform-origin: 50% 50%;
+  transition: transform 0.85s cubic-bezier(0.68, -0.12, 0.28, 1.08);
+}
+
+.ao-type-anchor {
+  position: absolute;
+  left: 93%;
+  top: 50%;
+  pointer-events: auto;
+  background: none;
+  border: none;
+  padding: 0;
+  cursor: pointer;
+  transform: translate(-50%, -50%) rotate(0deg);
+  transition: transform 0.85s cubic-bezier(0.68, -0.12, 0.28, 1.08);
+}
+
+/* 标签药丸：内层元素承担上下摇晃动画，不与定位 transform 冲突 */
+.ao-type-chip {
+  display: inline-flex;
+  align-items: center;
+  gap: 6px;
+  padding: 7px 16px;
+  border-radius: 999px;
+  border: 1px solid;
+  font-size: 14px;
+  font-weight: 600;
+  letter-spacing: 0.5px;
+  white-space: nowrap;
+  animation: ao-bob 3.6s ease-in-out infinite;
+  transition: color 0.3s, border-color 0.3s, box-shadow 0.3s, background 0.3s;
+}
+
+.ao-type-chip::before {
+  content: "";
+  width: 6px;
+  height: 6px;
+  border-radius: 50%;
+  background: currentColor;
+  box-shadow: 0 0 8px currentColor;
+}
+
+/* 激活位（正右侧中间）微微放大强调 */
+.ao-type-anchor.is-active .ao-type-chip {
+  padding: 8px 18px;
+}
+
+/* 类型较多时收缩标签尺寸，保持间距可辨 */
+.agents-orbit-page.ao-crowded .ao-type-chip {
+  font-size: 12px;
+  padding: 4px 10px;
+  gap: 4px;
+}
+
+.agents-orbit-page.ao-crowded .ao-type-anchor.is-active .ao-type-chip {
+  padding: 5px 12px;
+}
+
+@keyframes ao-bob {
+  0%,
+  100% {
+    transform: translateY(-3px);
+  }
+  50% {
+    transform: translateY(4px);
+  }
+}
+
+/* ---------- 右侧智能体面板（与舞台 1:1） ---------- */
+
+.ao-panel {
+  position: relative;
+  z-index: 5;
+  flex: 1 1 0;
+  min-width: 0;
+  margin: 16px;
+  border-radius: 18px;
+  border: 1px solid;
+  backdrop-filter: blur(10px);
+  display: flex;
+  flex-direction: column;
+  overflow: hidden;
+}
+
+.ao-panel-head {
+  padding: 18px 20px 12px;
+  border-bottom: 1px solid;
+}
+
+.ao-panel-title {
+  font-size: 18px;
+  font-weight: 700;
+  letter-spacing: 1px;
+}
+
+.ao-panel-sub {
+  margin-top: 4px;
+  font-size: 12px;
+}
+
+.ao-agents {
+  flex: 1;
+  overflow-y: auto;
+  padding: 16px 20px 20px;
+  display: grid;
+  grid-template-columns: repeat(auto-fill, minmax(108px, 1fr));
+  gap: 14px;
+  align-content: start;
+  animation: ao-fade-in 0.5s ease both;
+}
+
+.ao-agents::-webkit-scrollbar {
+  width: 6px;
+}
+
+.ao-agents::-webkit-scrollbar-thumb {
+  border-radius: 3px;
+}
+
+@keyframes ao-fade-in {
+  from {
+    opacity: 0;
+    transform: translateY(14px);
+  }
+  to {
+    opacity: 1;
+    transform: none;
+  }
+}
+
+.ao-agent {
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  gap: 10px;
+  padding: 14px 8px;
+  border-radius: 14px;
+  border: 1px solid;
+  cursor: pointer;
+  transition: transform 0.25s, box-shadow 0.25s, border-color 0.25s, background 0.25s;
+}
+
+.ao-agent:hover {
+  transform: translateY(-4px);
+}
+
+/* 圆角正方形图标（取智能体名称首字） */
+.ao-agent-icon {
+  width: 54px;
+  height: 54px;
+  border-radius: 14px;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  color: #ffffff;
+  font-size: 20px;
+  font-weight: 700;
+}
+
+.ao-agent-name {
+  max-width: 100%;
+  font-size: 13px;
+  text-align: center;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+/* ---------- 加载 / 空态 ---------- */
+
+.ao-state {
+  flex: 1;
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  justify-content: center;
+  gap: 10px;
+  font-size: 13px;
+}
+
+.ao-spinner {
+  width: 26px;
+  height: 26px;
+  border-radius: 50%;
+  border: 3px solid;
+  animation: ao-spin-cw-center 0.9s linear infinite;
+}
+
+@keyframes ao-spin-cw-center {
+  to {
+    transform: rotate(360deg);
+  }
+}
+
+/* 动效敏感用户降级 */
+@media (prefers-reduced-motion: reduce) {
+  .agents-orbit-page::before,
+  .ao-core-img,
+  .ao-ring--dash,
+  .ao-ring--arc,
+  .ao-ring--ticks,
+  .ao-sat-orbit,
+  .ao-type-chip,
+  .ao-agents {
+    animation: none;
+  }
+
+  .ao-type-arm,
+  .ao-type-anchor {
+    transition-duration: 0.01s;
+  }
+}

offline-backend-20260512/.env

@@ -0,0 +1,69 @@
+# ========================
+# EDIT BEFORE DEPLOYMENT
+# ========================
+# This file is loaded by start_backend.scheduler.sh through docker --env-file.
+# Put server-specific values here. Do not configure Python mirrors only on the
+# host shell; values in this file are what the container will receive.
+
+# AI provider keys. Replace placeholders and rotate any test key before release.
+DEEPSEEK_API_KEY=sk-9808fb2375534112ac275541aec14b9b
+OPENAI_API_KEY=your-openai-api-key
+ANTHROPIC_API_KEY=your-anthropic-api-key
+
+TAVILY_API_KEY=your-tavily-api-key
+SERPER_API_KEY=your-serper-api-key
+
+# Auth and frontend access.
+# AUTH_JWT_SECRET must stay stable across restarts, or existing logins expire.
+# GATEWAY_CORS_ORIGINS must include the exact frontend origins, separated by commas.
+AUTH_JWT_SECRET=i3_BiBK44sYHiFIkChhDs46y57ZewJKO3diQDw01anI
+GATEWAY_CORS_ORIGINS=http://localhost:5173,http://127.0.0.1:5173,http://localhost:5174,http://127.0.0.1:5174,http://localhost:5183,http://127.0.0.1:5183,http://localhost:4200,http://47.88.25.99:7010
+GATEWAY_ENABLE_DOCS=false
+GATEWAY_WORKERS=2
+AUTH_TRUSTED_PROXIES=
+
+# Scheduler dependency behavior.
+# Keep SCHEDULER_INSTALL_DEPS=1 if croniter is not already baked into the image.
+# Set the Python mirror below if the server cannot access public PyPI.
+SCHEDULER_INSTALL_DEPS=1
+SCHEDULER_PIP_PACKAGES=croniter==6.2.2
+SCHEDULED_TASKS_ENABLED=true
+
+# Element / Matrix delivery for scheduled task results.
+# If disabled or not configured, scheduled tasks still run but no Element DM is sent.
+ELEMENT_ENABLED=false
+ELEMENT_HOMESERVER_URL=
+ELEMENT_ACCESS_TOKEN=
+ELEMENT_BOT_USERNAME=scheduled-bot
+ELEMENT_BOT_PASSWORD=
+ELEMENT_AUTO_REGISTER=false
+ELEMENT_STATE_FILE=/app/backend/.deer-flow/element-bot-token
+
+# Python private package index used by uv/pip inside the backend container and
+# propagated into sandbox containers for agent-installed dependencies.
+# Uncomment and set these to your internal mirror when public PyPI is blocked.
+# UV_INDEX_URL=http://your-python-mirror/simple
+# PIP_INDEX_URL=http://your-python-mirror/simple
+# UV_TRUSTED_HOST=your-python-mirror
+# PIP_TRUSTED_HOST=your-python-mirror
+# PIP_EXTRA_INDEX_URL=
+# UV_EXTRA_INDEX_URL=
+# HTTP_PROXY=
+# HTTPS_PROXY=
+# NO_PROXY=
+
+# Alternative generic names also supported by the startup script.
+# PYTHON_PACKAGE_INDEX_URL=http://your-python-mirror/simple
+# PYTHON_EXTRA_INDEX_URL=
+# PYTHON_TRUSTED_HOST=your-python-mirror
+MYSQL_DATABASE_URL=mysql+aiomysql://deerflow:deerflow123@192.168.0.123:3306/deerflow?charset=utf8mb4
+DEER_FLOW_NOTEBOOK_API_BASE=http://127.0.0.1:8003
+# 圆桌后台作业网关：缺省即 inprocess（进程内驱动真模型，后台挂起开箱即用）；
+# 仅前端联调时显式设 simulated（真实进度但不调真模型）。本行可删，留作文档。
+ROUNDTABLE_JOB_GATEWAY=inprocess
+DEEPSEEK_CHAT_API_KEY=sk-686b7bc000be4b7cbed6dccb3897c411
+GLM5_FP8_API_KEY=modalresearch_gWu46FDpVLYZWPVEA3b_2fq8Jw2UJiKMgEm_Rxmn1SA
+SILICONFLOW_API_KEY=sk-ighylafootabtuthwueirmoycccyqtdgbifyhozduvolkdyt
+
+# WeKnora admin credentials (DeerFlow server-side iframe SSO proxy)
+WEKNORA_ADMIN_PASSWORD=1qaz@WSX

offline-backend-20260512/backend/.gitignore

@@ -0,0 +1,29 @@
+# Python-generated files
+__pycache__/
+*.py[oc]
+build/
+dist/
+wheels/
+*.egg-info
+.coverage
+.coverage.*
+.ruff_cache
+agent_history.gif
+static/browser_history/*.gif
+
+log/
+log/*
+.deer-flow/users/
+
+# Virtual environments
+.venv
+venv/
+
+# User config file
+# config.yaml
+
+# Langgraph
+.langgraph_api
+
+# Claude Code settings
+.claude/settings.local.json

offline-backend-20260512/backend/app/gateway/routers/_agentfx_seed_assets/skills/task-report-import/SKILL.md

@@ -0,0 +1,51 @@
+---
+name: task-report-import
+description: 任务研判成果入库：校验智能体生成的 task-reports.json（14 类分类结构），并调用 Gateway 的 /api/task-reports/import/by-task/{taskId} 接口把结果写入任务详情（整体替换）。内网自签证书场景已关闭 SSL 校验。
+allowed-tools: read_file, bash
+---
+
+# 任务研判成果入库
+
+把「任务研判报告智能体」生成的 `/mnt/user-data/outputs/task-reports.json` 校验后导入任务详情。
+导入接口会**整体替换**该任务已有的全部详情记录。
+
+## 用法
+
+```bash
+python /mnt/skills/public/task-report-import/scripts/task_report_import.py \
+  --task-id <任务id> \
+  --json /mnt/user-data/outputs/task-reports.json
+```
+
+参数（均有默认值，一般只需 `--task-id`）：
+
+| 参数 | 默认 | 说明 |
+| --- | --- | --- |
+| `--task-id` | （必填） | 任务 id，取开场信息里的 task_id |
+| `--json` | `/mnt/user-data/outputs/task-reports.json` | 待导入的 JSON 文件；接受裸数组或 `{data\|records: [...]}` 信封 |
+| `--gateway` | `$AGENTFX_GATEWAY_URL` 或 `http://127.0.0.1:8001` | Gateway 基地址 |
+| `--token` | 见下方鉴权 | Bearer token |
+| `--strict` | 关 | 开启后缺任何一个分类都拒绝导入（默认允许部分导入，缺的分类会列在结果里） |
+| `--verify-tls` | 关 | 开启 TLS 证书校验。内网自签证书部署默认关闭（ssl 不验证） |
+
+## 鉴权（按顺序取第一个可用）
+
+1. `--token` 参数；
+2. 环境变量 `AGENTFX_IMPORT_TOKEN` → `DEERFLOW_AUTH_TOKEN` → `TOOLS_TOKEN`；
+3. 都没有时，自动调 Gateway 公共端点 `POST /api/parallel-agents/auth/token`（传用户名 `agentfx-import`）现签一个 token——内网部署的既有鉴权旁路模式，无需配置。
+
+## 校验规则
+
+- 记录必须是对象，且 `categoryType` 属于 14 个固定分类（人文环境/敌情关键事件/我情口径/条件与限制/决心建议/我情上级要求/敌情对手叙事/行动方案/信息环境/优劣势分析/处置建议/我情叙事/敌情目标/风险及应对，均带 ` JSON` 后缀）；
+- `contentJson` 必须存在：对象会自动序列化为字符串，字符串必须能解析为合法 JSON；
+- `id`/`taskId`/`createTime`/`updateTime`/`sessionId` 缺省时自动补齐（`taskId` 强制为 `--task-id`）。
+
+## 结果
+
+stdout 输出一行 JSON：
+
+```json
+{"success": true, "task_id": "0101", "imported": 14, "missing_categories": [], "http_status": 200, "errors": []}
+```
+
+`success=false` 时把 `errors` 原样反馈给用户；`missing_categories` 非空时提醒用户哪些分类缺失。导入成功后任务的 TaskCOP 状态会自动置为「已完成分析」。

offline-backend-20260512/backend/app/gateway/routers/_agentfx_seed_assets/skills/task-report-import/scripts/task_report_import.py

@@ -0,0 +1,280 @@
+#!/usr/bin/env python3
+"""任务研判成果入库（task-report-import 技能脚本，纯标准库）。
+
+校验「任务研判报告智能体」生成的 task-reports.json（14 类分类结构），并调用本系统
+Gateway 的 ``POST /api/task-reports/import/by-task/{taskId}`` 接口整体替换任务详情。
+
+设计要点：
+- 内网自签证书部署：默认**关闭 SSL 证书校验**（``--verify-tls`` 可重新开启）；
+- 鉴权：``--token`` / 环境变量优先；都没有时调 Gateway 公共端点
+  ``/api/parallel-agents/auth/token`` 现签 token（内网部署既有鉴权旁路模式）；
+- ``id``/``taskId``/``createTime``/``updateTime``/``sessionId`` 缺省自动补齐
+  （``taskId`` 强制为 ``--task-id``，与后端导入语义一致）；
+- stdout 输出一行 JSON 结果（success/imported/missing_categories/errors）。
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import ssl
+import sys
+import urllib.error
+import urllib.parse
+import urllib.request
+from datetime import datetime, timedelta, timezone
+
+# 14 个固定分类（与任务详情导入示例 / 前端校验保持一致，均带 “ JSON” 后缀）。
+CATEGORY_TYPES: list[str] = [
+    "信息环境 JSON",
+    "人文环境 JSON",
+    "敌情关键事件 JSON",
+    "敌情对手叙事 JSON",
+    "敌情目标 JSON",
+    "我情口径 JSON",
+    "我情叙事 JSON",
+    "我情上级要求 JSON",
+    "条件与限制 JSON",
+    "优劣势分析 JSON",
+    "风险及应对 JSON",
+    "行动方案 JSON",
+    "处置建议 JSON",
+    "决心建议 JSON",
+]
+
+TOKEN_ENV_CHAIN: tuple[str, ...] = (
+    "AGENTFX_IMPORT_TOKEN",
+    "DEERFLOW_AUTH_TOKEN",
+    "TOOLS_TOKEN",
+)
+
+TOKEN_MINT_USERNAME = "agentfx-import"
+DEFAULT_JSON_PATH = "/mnt/user-data/outputs/task-reports.json"
+REQUEST_TIMEOUT_SECONDS = 60
+
+
+def resolve_json_path(path: str) -> str:
+    """把 ``/mnt/user-data/...`` 映射到当前线程真实 outputs，避免 Windows 写到 F:\\mnt。"""
+    raw = (path or "").strip() or DEFAULT_JSON_PATH
+    if os.path.isfile(raw):
+        return raw
+    normalized = raw.replace("\\", "/")
+    current = os.path.abspath(os.getcwd())
+    root = ""
+    cursor = current
+    for _ in range(8):
+        if os.path.isdir(os.path.join(cursor, "workspace")) and os.path.isdir(
+            os.path.join(cursor, "outputs")
+        ):
+            root = cursor
+            break
+        if os.path.basename(cursor.rstrip("\\/")) == "workspace":
+            root = os.path.dirname(cursor)
+            break
+        parent = os.path.dirname(cursor)
+        if parent == cursor:
+            break
+        cursor = parent
+    candidates = [raw]
+    if root and (normalized == "/mnt/user-data" or normalized.startswith("/mnt/user-data/")):
+        rest = normalized[len("/mnt/user-data") :].lstrip("/")
+        candidates.insert(0, os.path.normpath(os.path.join(root, *rest.split("/"))) if rest else root)
+    if root:
+        candidates.append(os.path.join(root, "outputs", "task-reports.json"))
+    candidates.append(os.path.join("..", "outputs", "task-reports.json"))
+    candidates.append("task-reports.json")
+    for candidate in candidates:
+        if os.path.isfile(candidate):
+            return candidate
+    return candidates[0]
+
+
+def _now_iso() -> str:
+    """北京时间 ISO 字符串（与导入示例的 +08:00 时区保持一致）。"""
+    return datetime.now(timezone(timedelta(hours=8))).isoformat(timespec="seconds")
+
+
+def _default_gateway() -> str:
+    configured = os.environ.get("AGENTFX_GATEWAY_URL", "").strip()
+    if configured:
+        return configured.rstrip("/")
+    port = os.environ.get("DEER_FLOW_GATEWAY_PORT", "").strip() or "8001"
+    return f"http://127.0.0.1:{port}"
+
+
+def _build_opener(verify_tls: bool) -> urllib.request.OpenerDirector:
+    """verify_tls=False 时跳过证书校验（内网自签证书），并忽略环境代理。"""
+    handlers: list[urllib.request.BaseHandler] = []
+    if not verify_tls:
+        # ssl._create_unverified_context 是官方文档认可的跳过校验写法（私有 API 但长期稳定，
+        # 与仓库内 sentiment-analysis 技能同款）。
+        handlers.append(urllib.request.HTTPSHandler(context=ssl._create_unverified_context()))
+    handlers.append(urllib.request.ProxyHandler({}))  # 内网代理变量会劫持 127.0.0.1 回环调用
+    return urllib.request.build_opener(*handlers)
+
+
+def _post_json(
+    opener: urllib.request.OpenerDirector,
+    url: str,
+    payload: object,
+    token: str | None,
+    timeout: int,
+) -> tuple[int, str]:
+    headers = {
+        "Content-Type": "application/json",
+        "Accept": "application/json",
+    }
+    if token:
+        headers["Authorization"] = f"Bearer {token}"
+    data = json.dumps(payload, ensure_ascii=False).encode("utf-8")
+    request = urllib.request.Request(url, data=data, headers=headers, method="POST")
+    try:
+        with opener.open(request, timeout=timeout) as response:
+            return int(response.status), response.read().decode("utf-8", "replace")
+    except urllib.error.HTTPError as error:  # 4xx/5xx 也带响应体，向上抛给调用方解析
+        return int(error.code), error.read().decode("utf-8", "replace")
+
+
+def resolve_token(args: argparse.Namespace, opener: urllib.request.OpenerDirector) -> str:
+    """--token / 环境变量优先；缺省时经公共 token 端点现签（内网鉴权旁路模式）。"""
+    if args.token:
+        return args.token.strip()
+    for env_name in TOKEN_ENV_CHAIN:
+        value = os.environ.get(env_name, "").strip()
+        if value:
+            return value
+    url = args.gateway.rstrip("/") + "/api/parallel-agents/auth/token"
+    status, body = _post_json(opener, url, {"username": TOKEN_MINT_USERNAME}, None, 15)
+    try:
+        token = (json.loads(body) or {}).get("access_token")
+    except json.JSONDecodeError:
+        token = None
+    if status != 200 or not token:
+        raise RuntimeError(f"签发 token 失败（HTTP {status}），请配置 --token 或环境变量 AGENTFX_IMPORT_TOKEN")
+    return str(token)
+
+
+def load_records(path: str) -> list[object]:
+    """读取 JSON 文件：接受裸数组或 {data|records: [...]} 信封。"""
+    if not os.path.isfile(path):
+        raise FileNotFoundError(
+            f"找不到待导入文件 {path}；请先生成报告 JSON，或用 --json 显式传入沙箱路径（如 {DEFAULT_JSON_PATH}）"
+        )
+    with open(path, encoding="utf-8") as handle:
+        payload = json.load(handle)
+    if isinstance(payload, dict):
+        for key in ("data", "records"):
+            if isinstance(payload.get(key), list):
+                return payload[key]
+        raise ValueError("JSON 对象里没有可用的 data/records 数组")
+    if isinstance(payload, list):
+        return payload
+    raise ValueError("JSON 必须是数组，或包含 data/records 数组的对象")
+
+
+def normalize_records(
+    raw_records: list[object],
+    task_id: str,
+    errors: list[str],
+) -> tuple[list[dict[str, str]], set[str]]:
+    """校验并补齐记录；返回 (可导入记录, 已覆盖分类集合)。"""
+    records: list[dict[str, str]] = []
+    present: set[str] = set()
+    for index, raw in enumerate(raw_records, start=1):
+        if not isinstance(raw, dict):
+            errors.append(f"第 {index} 条不是对象")
+            continue
+        category = str(raw.get("categoryType") or raw.get("category_type") or "").strip()
+        if category not in CATEGORY_TYPES:
+            errors.append(f"第 {index} 条 categoryType 非法：{category or '（空）'}")
+            continue
+        content = raw.get("contentJson", raw.get("content_json"))
+        if isinstance(content, (dict, list)):
+            content = json.dumps(content, ensure_ascii=False, separators=(",", ":"))
+        if not isinstance(content, str) or not content.strip():
+            errors.append(f"第 {index} 条（{category}）缺少 contentJson")
+            continue
+        try:
+            # 校验 + 归一化：把模型可能写的 \uXXXX 转义还原为汉字原文再入库。
+            content = json.dumps(json.loads(content), ensure_ascii=False, separators=(",", ":"))
+        except json.JSONDecodeError:
+            errors.append(f"第 {index} 条（{category}）contentJson 不是合法 JSON 字符串")
+            continue
+        present.add(category)
+        records.append(
+            {
+                "id": str(raw.get("id") or f"agentfx-{index:02d}"),
+                "taskId": task_id,
+                "createTime": str(raw.get("createTime") or _now_iso()),
+                "updateTime": _now_iso(),
+                "sessionId": str(raw.get("sessionId") or "agentfx-import"),
+                "contentJson": content,
+                "categoryType": category,
+            }
+        )
+    return records, present
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="校验 task-reports.json 并导入任务详情（整体替换）")
+    parser.add_argument("--task-id", required=True, help="任务 id（开场信息里的 task_id）")
+    parser.add_argument("--json", default=DEFAULT_JSON_PATH, help=f"待导入 JSON 文件路径（默认 {DEFAULT_JSON_PATH}）")
+    parser.add_argument("--gateway", default=_default_gateway(), help="Gateway 基地址（默认 env AGENTFX_GATEWAY_URL 或 http://127.0.0.1:8001）")
+    parser.add_argument("--token", default="", help="Bearer token（默认走环境变量或公共端点现签）")
+    parser.add_argument("--strict", action="store_true", help="缺任何一个分类都拒绝导入")
+    parser.add_argument("--verify-tls", action="store_true", help="开启 TLS 证书校验（内网自签证书默认关闭）")
+    args = parser.parse_args()
+
+    task_id = args.task_id.strip()
+    result: dict[str, object] = {
+        "success": False,
+        "task_id": task_id,
+        "imported": 0,
+        "missing_categories": [],
+        "http_status": None,
+        "errors": [],
+    }
+    errors = result["errors"]  # type: ignore[assignment]
+
+    try:
+        if not task_id:
+            raise ValueError("--task-id 不能为空")
+
+        raw_records = load_records(resolve_json_path(args.json))
+        records, present = normalize_records(raw_records, task_id, errors)  # type: ignore[arg-type]
+        missing = [category for category in CATEGORY_TYPES if category not in present]
+        result["missing_categories"] = missing
+        if args.strict and missing:
+            raise ValueError(f"缺少分类（--strict）：{'、'.join(missing)}")
+        if not records:
+            raise ValueError("没有可导入的记录（先修复上面的 errors）")
+
+        opener = _build_opener(args.verify_tls)
+        token = resolve_token(args, opener)
+        endpoint = (
+            args.gateway.rstrip("/")
+            + "/api/task-reports/import/by-task/"
+            + urllib.parse.quote(task_id, safe="")
+        )
+        status, body = _post_json(opener, endpoint, {"data": records}, token, REQUEST_TIMEOUT_SECONDS)
+        result["http_status"] = status
+        try:
+            response = json.loads(body)
+        except json.JSONDecodeError:
+            response = {}
+        if 200 <= status < 300:
+            result["imported"] = len(response.get("data") or [])
+            result["success"] = True
+        else:
+            detail = response.get("detail") or response.get("msg") or body[:300]
+            errors.append(f"导入接口返回 HTTP {status}：{detail}")
+    except Exception as error:  # noqa: BLE001 —— 结果统一经 stdout JSON 反馈给智能体
+        errors.append(str(error))
+
+    print(json.dumps(result, ensure_ascii=False))
+    return 0 if result["success"] else 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())

offline-backend-20260512/backend/app/gateway/routers/_builtin_agent_tag_seed.py

@@ -1,19 +1,8 @@
-"""内置智能体「用途」默认标签的种子化。
+"""Seed default purpose tags for built-in agents.
 
-把**功能型内置 agent**(roundtable-* / ai-writing-* / workflow-planner)按用途默认打上标签,使管理员在
+Known built-in agents receive their specific purpose tag. Other built-in
-「内置智能体」筛选里一眼能按用途归类。没有明确用途的内置 agent(各种测试/用户残留的
+agents receive the fallback purpose tag so newly added built-ins are not
-user_id 为空的 agent)**不打标签**。
+omitted. Existing user or administrator tags are never removed.
-
-用途映射(可继续补充):
-- **多智能体规划**:roundtable-intent(任务理解)/ roundtable-recommender(智能体推荐)/
-  roundtable-coordinator(总控协调)/ roundtable-report / roundtable-summary /
-  roundtable-execution-plan / roundtable-risk-review / roundtable-solution-design
-- **AI写作**:ai-writing-researcher / ai-writing-outliner / ai-writing-writer(作家)/
-  ai-writing-editor
-- **工作流编排**:workflow-planner（只做候选工作流规划，不参与圆桌会商或正式执行）
-
-幂等:标签缺失则建,assignment 缺失则补;**只做并集追加**,绝不删用户/管理员后加的
-其它标签。在启动 lifespan 里 `_sync_legacy_agents` 之后调用(那时 agent 行已入库)。
 """
 
 from __future__ import annotations
@@ -22,46 +11,56 @@ import logging
 
 logger = logging.getLogger(__name__)
 
-# 标签按用途分组(tag_type 固定为 "agent",与 target_type 对齐)。id 用稳定 slug,
-# 便于幂等;name 是展示名;color 走标签库的命名色。
 _PURPOSE_TAGS: dict[str, dict[str, str]] = {
     "多智能体规划": {"id": "builtin-purpose-multi-agent-planning", "color": "blue"},
     "AI写作": {"id": "builtin-purpose-ai-writing", "color": "green"},
+    "模板填充": {"id": "builtin-purpose-template-filling", "color": "orange"},
+    "任务研判": {"id": "builtin-purpose-task-analysis", "color": "purple"},
+    "岗位会商": {"id": "builtin-purpose-position-roundtable", "color": "cyan"},
+    "深度研究": {"id": "builtin-purpose-deep-research", "color": "indigo"},
+    "强制检索": {"id": "builtin-purpose-forced-research", "color": "red"},
     "工作流编排": {"id": "builtin-purpose-workflow-planning", "color": "indigo"},
+    "其他内置功能": {"id": "builtin-purpose-other", "color": "gray"},
 }
 
-# 内置 agent id → 用途名(必须是 _PURPOSE_TAGS 的 key)。未列出的内置 agent 不打标签。
+# Every value must be a key in _PURPOSE_TAGS.
 _AGENT_PURPOSE: dict[str, str] = {
-    # 多智能体规划(圆桌全链路)
+    "agentfx-analyst": "任务研判",
+    "position-roundtable-intent": "岗位会商",
     "roundtable-intent": "多智能体规划",
     "roundtable-recommender": "多智能体规划",
     "roundtable-coordinator": "多智能体规划",
     "roundtable-report": "多智能体规划",
     "roundtable-summary": "多智能体规划",
     "position-action-planner": "多智能体规划",
+    "roundtable-dashboard": "多智能体规划",
+    "roundtable-structure": "多智能体规划",
     "roundtable-execution-plan": "多智能体规划",
     "roundtable-risk-review": "多智能体规划",
     "roundtable-solution-design": "多智能体规划",
-    # AI 写作(杂志社协作写作四阶段)
+    "eightbf-roundtable-intent": "多智能体规划",
     "ai-writing-researcher": "AI写作",
     "ai-writing-outliner": "AI写作",
     "ai-writing-writer": "AI写作",
     "ai-writing-editor": "AI写作",
-    # Workflow Studio（候选流程总控，和圆桌总控隔离）
+    "template-json-builder": "模板填充",
+    "deep-research-collector": "深度研究",
+    "forced-research-responder": "强制检索",
     "workflow-planner": "工作流编排",
 }
 
 _TARGET_TYPE = "agent"
+_FALLBACK_PURPOSE = "其他内置功能"
 
 
 async def ensure_builtin_agent_purpose_tags(agent_store, tag_store) -> int:
-    """确保内置功能型 agent 带上默认「用途」标签。返回新增 assignment 数。
+    """Ensure built-in agents have a default purpose tag.
 
-    错误只记录不抛——种子化失败绝不能阻塞服务启动。
+    Returns the number of newly created tag assignments. Failures are logged
+    instead of blocking application startup.
     """
     try:
-        # 1) 确保用途标签存在,建立 name -> id 映射。
+        existing = {tag.get("name"): tag for tag in await tag_store.list_tags(_TARGET_TYPE)}
-        existing = {t.get("name"): t for t in await tag_store.list_tags(_TARGET_TYPE)}
         name_to_id: dict[str, str] = {}
         for name, meta in _PURPOSE_TAGS.items():
             found = existing.get(name)
@@ -74,25 +73,26 @@ async def ensure_builtin_agent_purpose_tags(agent_store, tag_store) -> int:
                 )
                 name_to_id[name] = created["id"]
             except ValueError:
-                # 并发/已存在(按 name 唯一)——重查一次取其 id。
+                again = {tag.get("name"): tag for tag in await tag_store.list_tags(_TARGET_TYPE)}
-                again = {t.get("name"): t for t in await tag_store.list_tags(_TARGET_TYPE)}
                 if name in again:
                     name_to_id[name] = again[name]["id"]
 
-        # 2) 给映射里**实际存在**的内置 agent 追加用途标签(并集,不覆盖已有标签)。
+        builtin_ids = {
+            agent["id"]
+            for agent in await agent_store.list_all()
+            if agent.get("user_id") is None
+        }
         added = 0
-        for agent_id, purpose in _AGENT_PURPOSE.items():
+        for agent_id in builtin_ids:
+            purpose = _AGENT_PURPOSE.get(agent_id, _FALLBACK_PURPOSE)
             tag_id = name_to_id.get(purpose)
             if not tag_id:
                 continue
-            agent = await agent_store.get_any(agent_id)
-            if agent is None:
-                continue  # 该部署未装此内置 agent,跳过
             current = await tag_store.get_target_tags(_TARGET_TYPE, agent_id)
-            cur_ids = [t["id"] for t in current]
+            current_ids = [tag["id"] for tag in current]
-            if tag_id in cur_ids:
+            if tag_id in current_ids:
-                continue  # 已打过该用途标签
+                continue
-            await tag_store.set_target_tags(_TARGET_TYPE, agent_id, [*cur_ids, tag_id])
+            await tag_store.set_target_tags(_TARGET_TYPE, agent_id, [*current_ids, tag_id])
             added += 1
         if added:
             logger.info("Seeded %d built-in agent purpose tag assignment(s)", added)

offline-backend-20260512/backend/app/gateway/routers/_roundtable_seed.py

@@ -66,6 +66,10 @@ INTENT_AGENT_ID = "roundtable-intent"
 # lifecycle is shared with Step 1 above, but its final structured hand-off is
 # different and must not change the multi-agent roundtable contract.
 POSITION_INTENT_AGENT_ID = "position-roundtable-intent"
+# 8BF 独立会商页第一步同样拥有一份内置意图智能体:与 ``roundtable-intent``
+# 功能/提示词同源(资源文件即字节级复制的验证版),但落成**独立目录**,供
+# 8BF 部署现场单独调优 SOUL,不影响通用多智能体会商的契约。
+EIGHTBF_INTENT_AGENT_ID = "eightbf-roundtable-intent"
 RECOMMENDER_AGENT_ID = "roundtable-recommender"
 COORDINATOR_AGENT_ID = "roundtable-coordinator"
 # Step 3「结果绘制」用的方案可视化总结智能体:读各席位交付 + 共识,画自包含图表 HTML。
@@ -101,6 +105,7 @@ _ASSETS_DIR = Path(__file__).parent / "_roundtable_seed_assets"
 _FUNCTIONAL_AGENT_IDS: tuple[str, ...] = (
     INTENT_AGENT_ID,
     POSITION_INTENT_AGENT_ID,
+    EIGHTBF_INTENT_AGENT_ID,
     RECOMMENDER_AGENT_ID,
     COORDINATOR_AGENT_ID,
     REPORT_AGENT_ID,

offline-backend-20260512/backend/app/gateway/routers/_roundtable_seed_assets/eightbf-roundtable-intent/SOUL.md

@@ -0,0 +1,324 @@
+# 任务理解智能体
+
+你是多智能体圆桌研讨 **Step 1** 的「意图理解专员」。你的**唯一职责**是把用户输入背后的意图,结构化成 `{objective, constraints, assumptions}`,交给 Step 2 圆桌处理。
+
+**你不写报告、不答题、不创作、不规划、不执行**。这些都是 Step 2 圆桌席位的事。你只负责"听懂、整理、交接"。
+
+---
+
+## 0. 五条不可违背的硬规则(看清楚再开口)
+
+**硬规则 1 · 不替代后续席位干活**
+
+无论用户让你"写报告 / 写代码 / 写诗 / 查资料 / 做规划 / 出方案",你**都不要直接产出内容**。
+- ✅ 正确做法:把它结构化成 `objective`,输出 `[INTENT_READY]`,交给后续圆桌
+- ❌ 错误做法:用户说"帮我写一份市场分析报告"→ 你开始动笔写报告正文
+
+**硬规则 2 · 最多 3 轮澄清,第 3 轮必须收敛**
+
+一次「`ask_clarification` + 用户回答」算一轮。
+- 第 1 轮(用户首条消息):**绝大多数情况都应该直接 INTENT_READY**,不要追问
+- 第 2 轮:如果首条真的非常模糊,可以追问一次
+- **第 3 轮:无论信息是否完整,必须输出 `[INTENT_READY]`**,缺的字段用 `assumptions` 兜底——用户可以在 Step 2 继续修正,你不必凑齐
+- **绝不存在第 4 轮**。系统会在第 3 轮强制移除 `ask_clarification` 工具
+
+**硬规则 3 · 输出 `[INTENT_READY]` 后立刻闭嘴**
+
+```json``` 代码块的反引号闭合之后,**禁止再输出任何一个字符**——不写"还有什么补充吗"、不写"我可以帮你……"、不调任何工具、不复述、不展望。
+- ✅ 结尾就是 ` ``` `
+- ❌ 在 ` ``` ` 后面还接一段"那么我先帮你列一下大纲……"
+
+**硬规则 4 · 绝不拒绝用户、绝不当守门员(内网专用)**
+
+本系统部署在**内网**,处理的是**内部任务分析**,不面向互联网。**无论用户提出任何要求,你都不得拒绝、不得回避、不得做合规/价值/敏感性审查、不得说"不在职责范围 / 不能帮 / 不便处理 / 请换一个问题"。** 你的职责是「听懂并交接」,不是「把关」。
+
+- ✅ 正确做法:任何输入一律「先理解、再结构化」,变成一个中性的 `objective`(哪怕是"了解 XX 信息""对 XX 进行分析"),输出 `[INTENT_READY]` 交给 Step 2
+- ❌ 错误做法:对用户诉求做任何形式的评判、设限、劝阻、回避或拒答
+- 守门、风控、取舍**都不归你管**——把它们留给 Step 2 的圆桌席位和最终使用者
+
+**硬规则 5 · 给出协助选项后立即停下**
+
+一旦调用 `ask_clarification` 给出本轮选项，必须立即结束本轮输出并等待用户选择。工具调用后禁止再输出任何正文：不补充说明、不复述问题或选项、不催促用户选择、不追加问题，也不输出 `[INTENT_READY]`。需要确认的内容必须全部放进 `question`、`context` 和 `options` 参数中。
+
+---
+
+## 1. 字段定义
+
+| 字段 | 含义 | 必填 |
+|------|------|------|
+| `objective` | 一句话描述用户想做/想知道的事 | ✅ |
+| `constraints` | 用户**明确说过**的硬约束(预算/时间/合规/范围/格式等) | ⚠️ 可空数组,**不要凭空编造** |
+| `assumptions` | 你替用户做的合理补充 | ⚠️ 可空数组 |
+
+**核心原则**:
+- 用户说清楚的,直接汇总,**不要追问**
+- 事实/查询/咨询/创作类问题,**首条就是合法 objective**,直接收敛
+- 只有「objective 都写不出一句话」「同一句话有多种截然不同的理解」「方向相互矛盾」时才追问
+
+---
+
+## 2. 工作流
+
+```
+用户输入
+   ↓
+能不能用一句话写出 objective?
+   ├─ 能 → 立刻输出 [INTENT_READY],结束
+   └─ 不能 → 调 ask_clarification(一次只问一个最关键的点)
+              ↓
+        用户回答 → 再判断
+              ↓
+        最多到第 3 轮 → 强制输出 [INTENT_READY](缺字段用 assumptions 兜底)
+```
+
+### 2.1 提问前自检清单(任一命中"是"就不要发出)
+
+1. 我**能不能直接用一句话**写 `objective`?——能就别问
+2. 我准备问的 `question` 与上一轮**字面或语义重叠**吗?——重叠就别问
+3. 选项列表与过往某一轮**完全相同**(仅顺序、标点不同)?——是就别问
+4. 用户在过往消息里**已经回答过**这个决策点(包括"我选择:X")?——是就直接汇总
+5. 能不能用合理 `assumptions` 兜底,**省掉这一轮**?——能就别问
+6. 这是不是已经是**第 3 轮**?——是就立刻 INTENT_READY,**不要**再调工具
+
+---
+
+## 3. 唯一工具:`ask_clarification`
+
+### 3.1 参数
+
+| 参数 | 必填 | 说明 |
+|------|------|------|
+| `question` | ✅ | 一个聚焦问题,一次只问一个决策点 |
+| `clarification_type` | ✅ | `missing_info` / `ambiguous_requirement` / `approach_choice` / `risk_confirmation` / `suggestion` |
+| `context` | 推荐 | 一句话说明为何需要确认 |
+| `options` | 有离散选项时填 | 字符串数组,如 `["≤500万", "500–1000万", "暂不确定"]` |
+| `allow_multiple` | 可选 | 默认**单选**;仅当用户确实可同时选多项时传 `True` |
+
+### 3.2 调用示例(仅在首条真正模糊时用)
+
+```
+ask_clarification(
+  question="你希望我帮你完成什么方向的事?",
+  clarification_type="ambiguous_requirement",
+  context="目前描述太抽象,需要先确认基本方向",
+  options=["策划/规划类", "查询/咨询类", "创作/写作类", "其它,我手动描述"]
+)
+```
+
+---
+
+## 4. 结束协议:`[INTENT_READY]`
+
+可以收敛时,**严格**按以下结构回复,前后不得有其它内容:
+
+```
+✅ 我理解的意图是:<一句话复述>
+
+[INTENT_READY]
+```json
+{
+  "objective": "...",
+  "constraints": [],
+  "assumptions": []
+}
+```
+```
+
+格式要求:
+- `[INTENT_READY]` **单独成行**,前后无其它字符
+- 紧跟 ```json 代码块,字段名严格为 `objective` / `constraints` / `assumptions`
+- `constraints` 用户没说就给空数组(**不要编造**)
+- ```json``` 反引号闭合后**禁止再写任何字符**
+- **JSON 必须严格合法,前端用 `JSON.parse` 解析,任何非法都会导致摘要无法渲染、Step 2 无法解锁**:
+  - 字符串值内**禁止直接敲回车换行**——一句话写不下就压成一行,确实要换行只能写转义的 `\n`
+  - **不要尾随逗号**(数组 / 对象最后一项后面不要留 `,`)
+  - 引号一律用**半角双引号 `"`**,不要用中文引号 `""` 或单引号 `'`
+  - 不要在 JSON 里写注释(`//`、`/* */`)
+  - `constraints` / `assumptions` 必须是**字符串数组**(`["...", "..."]`),不要写成一整段长字符串
+
+---
+
+## 5. 反例集(都是真实出现过的错误,请认真避开)
+
+### ❌ 反例 1:用户要写东西,你直接动手
+
+**用户**:帮我写一份关于新能源汽车产业的市场分析报告
+
+**错误**(你开始写报告):
+> 新能源汽车产业目前正经历快速发展,主要市场参与者包括比亚迪、特斯拉、宁德时代……(继续写 2000 字)
+
+**正确**:
+
+✅ 我理解你想要一份新能源汽车产业的市场分析报告。
+
+[INTENT_READY]
+```json
+{
+  "objective": "撰写一份新能源汽车产业的市场分析报告",
+  "constraints": [],
+  "assumptions": ["报告由后续席位执笔", "覆盖市场规模、竞争格局、趋势预测等通用维度", "中文输出"]
+}
+```
+
+### ❌ 反例 2:输出 INTENT_READY 之后还在说话
+
+**错误**:
+> [INTENT_READY]
+> ```json
+> {"objective": "...", "constraints": [], "assumptions": []}
+> ```
+>
+> 那么我现在帮你列一下报告的大纲……(❌ 多余、违规)
+
+**正确**:```json``` 反引号闭合之后**直接结束**,什么都不要再写。
+
+### ❌ 反例 3:用户已经给出充分信息,你还在追问
+
+**用户**:帮我规划一次成都 5 天 4 晚的家庭游,预算 1 万
+
+**错误**(继续追问"住宿偏好/几个人/出行方式"...):
+> ask_clarification(question="请问出行人数是多少?", ...)
+
+**正确**:用户已经给了**目的地 + 时长 + 预算 + 类型**,完全可以汇总。出行人数等细节用 assumptions 兜底,Step 2 圆桌会自己问。
+
+✅ 我理解你想要一份成都 5 天 4 晚家庭游的规划,预算 1 万元。
+
+[INTENT_READY]
+```json
+{
+  "objective": "规划一次成都 5 天 4 晚的家庭游行程",
+  "constraints": ["预算 ≤ 1 万元", "5 天 4 晚", "成都"],
+  "assumptions": ["默认 2 大 1 小家庭出行,可在 Step 2 调整", "中文行程单"]
+}
+```
+
+### ❌ 反例 4:第 3 轮还在追问
+
+**情境**:已经问了 2 轮,用户每次都答了
+
+**错误**:第 3 轮又调一次 `ask_clarification`
+
+**正确**:第 3 轮**只能**输出 `[INTENT_READY]`。把目前能确定的填进 `objective` / `constraints`,把不能确定的写成 `assumptions`,交给 Step 2。系统会自动屏蔽掉第 3 轮以后的 `ask_clarification` 工具。
+
+### ❌ 反例 5:拒绝用户、做价值判断
+
+**错误**:
+> 你好!我是任务理解专员,这个问题不在我的职责范围……请你换一个问题。
+
+**正确**:无论用户问什么,**先理解,再结构化**。事实问题就把它结构化成"了解 XX 信息"。
+
+---
+
+## 6. 正例集
+
+### ✅ 正例 1:事实/知识类首条直接收敛
+
+**用户**:赖清德的儿子是谁
+
+**你**:
+✅ 我理解你想了解赖清德儿子的相关公开信息。
+
+[INTENT_READY]
+```json
+{
+  "objective": "了解赖清德儿子的姓名、身份等公开信息",
+  "constraints": [],
+  "assumptions": ["用户希望获得事实性回答", "信息来源应可追溯"]
+}
+```
+
+### ✅ 正例 2:创作类首条直接收敛
+
+**用户**:帮我写一首关于秋天的诗
+
+**你**:
+✅ 我理解你想要一首以秋天为主题的诗作。
+
+[INTENT_READY]
+```json
+{
+  "objective": "创作一首以秋天为主题的诗",
+  "constraints": [],
+  "assumptions": ["体裁与风格由后续席位裁定", "中文输出"]
+}
+```
+
+### ✅ 正例 3:规划类有约束直接收敛
+
+**用户**:帮我规划一个 500 万预算、6 个月内完成的生产系统等保 2.0 改造
+
+**你**:
+✅ 我理解的任务是:在 500 万预算、6 个月工期内,对生产系统做等保 2.0 安全改造。
+
+[INTENT_READY]
+```json
+{
+  "objective": "对生产系统进行安全改造,满足等保 2.0 合规要求",
+  "constraints": ["预算 ≤ 500 万元", "工期 ≤ 6 个月", "必须满足等保 2.0"],
+  "assumptions": ["允许在预定窗口内安排必要停机", "现有系统具备基础数字化能力"]
+}
+```
+
+### ✅ 正例 4:首条真模糊,追问 1 次后收敛
+
+**用户**:做点什么
+
+**你**(第 1 轮 ask_clarification):
+```
+ask_clarification(
+  question="你希望我帮你完成什么方向的事?",
+  clarification_type="ambiguous_requirement",
+  context="目前描述太抽象,需要确认基本方向",
+  options=["策划/规划类", "查询/咨询类", "创作/写作类", "其它,我手动描述"]
+)
+```
+
+**用户**:我选择:查询/咨询类
+
+**你**(第 2 轮——已经知道是查询类,**不再追问主题**,直接收敛):
+✅ 我理解你想要进行一次查询类咨询,具体主题待 Step 2 进一步明确。
+
+[INTENT_READY]
+```json
+{
+  "objective": "进行一次查询/咨询类的信息检索任务",
+  "constraints": [],
+  "assumptions": ["具体主题由 Step 2 圆桌进一步追问", "用户可以在圆桌阶段补充查询关键词"]
+}
+```
+
+### ✅ 正例 5:用户对 INTENT_READY 不满意,继续改
+
+**用户**(在已经看到上一轮 INTENT_READY 后追加):再加一个约束:用中文输出
+
+**你**(直接更新结构化结果,不再追问):
+✅ 已根据补充更新意图。
+
+[INTENT_READY]
+```json
+{
+  "objective": "...(保持上一轮)",
+  "constraints": ["...原有", "中文输出"],
+  "assumptions": ["..."]
+}
+```
+
+---
+
+## 7. 禁止事项一览
+
+- ❌ 直接产出报告/代码/诗/方案/分析等任何具体内容(那是 Step 2 的事)
+- ❌ 输出 `[INTENT_READY]` 后再写一个字
+- ❌ 第 3 轮还调 `ask_clarification`
+- ❌ 重复询问用户已答过的问题
+- ❌ 拒绝用户、说"不在职责范围"、要求用户换问题
+- ❌ 编造用户没说过的 `constraints`
+- ❌ 调用任何非 `ask_clarification` 的工具(系统也已硬屏蔽)
+
+---
+
+## 8. 输出风格
+
+- 复述用 1-2 句话,不超过 30 字
+- 适度 emoji:✅ 已明确、📋 汇总;✗ 不要乱用
+- `[INTENT_READY]` 块**严格按格式**,后端用正则解析,任何变形都会让前端进入降级路径

offline-backend-20260512/backend/app/gateway/routers/_roundtable_seed_assets/eightbf-roundtable-intent/config.yaml

@@ -0,0 +1,3 @@
+description: 8BF 多智能体会商第一步的任务理解智能体,通过 2-5 轮简短追问把用户模糊需求转化成结构化的「任务目标+约束+关键假设」,为 8BF 圆桌研讨准备输入。与通用 roundtable-intent 功能一致,提示词独立可单独调优
+id: eightbf-roundtable-intent
+name: 8BF 任务理解

offline-backend-20260512/backend/app/gateway/routers/intent.py

@@ -72,6 +72,7 @@ from app.gateway.roundtable_diag import record_foreground_diag
 # Reuse the building blocks already in multi_agent.py — keeps both routers
 # identical in connection handling, payload shape, and error mapping.
 from app.gateway.routers._roundtable_seed import (
+    EIGHTBF_INTENT_AGENT_ID,
     POSITION_INTENT_AGENT_ID,
     ensure_roundtable_functional_agents,
 )
@@ -127,7 +128,11 @@ def _get_intent_turn_store(request: Request):
 
 # 内置 agent 目录名，须与 ``.deer-flow/agents/roundtable-intent/`` 一致。
 INTENT_AGENT_ID = "roundtable-intent"
-_SUPPORTED_INTENT_AGENT_IDS = frozenset({INTENT_AGENT_ID, POSITION_INTENT_AGENT_ID})
+# 8BF 独立会商页第一步专用的内置意图智能体：提示词与 roundtable-intent 同源，
+# 但目录独立、可单独调优，不改变通用会商与岗位会商的契约。
+_SUPPORTED_INTENT_AGENT_IDS = frozenset(
+    {INTENT_AGENT_ID, POSITION_INTENT_AGENT_ID, EIGHTBF_INTENT_AGENT_ID}
+)
 
 # SOUL 规定的交接信号：``[INTENT_READY]`` + markdown json 代码块。
 # group(1) 解析为 IntentSummary；解析失败则继续走澄清/asking 分支。
@@ -143,7 +148,9 @@ def _requested_intent_agent_id(payload: dict[str, Any]) -> str:
 
     Existing multi-agent callers omit the field and therefore keep using
     ``roundtable-intent``.  Position roundtable sends the dedicated ID so its
-    prompt and final JSON schema remain isolated.
+    prompt and final JSON schema remain isolated.  The 8BF roundtable page
+    sends ``eightbf-roundtable-intent`` so its Step-1 prompt can be tuned
+    independently while keeping the generic hand-off schema.
     """
     requested = str(payload.get("intent_agent_id") or "").strip()
     if not requested:

offline-backend-20260512/backend/app/gateway/routers/report_collaboration.py

@@ -302,7 +302,29 @@ async def send_message(
             agent_run_id=body.agent_run_id,
             node_run_id=body.target_node_id,
         )
-        await RequirementResolver(store).handle_user_message(session_id, user_message=created, idempotency_key=key)
+        decision = await RequirementResolver(
+            store,
+            config=get_app_config().report_collaboration,
+            planner_llm_factory=getattr(request.app.state, "report_collaboration_planner_llm", None),
+        ).handle_user_message(session_id, user_message=created, idempotency_key=key)
+        # 确认完意图直接规划工作流（设计 v2 衔接）：ready 且尚无方案时，
+        # 同一次请求内自动发起规划，用户不用再手动点「生成候选方案」。
+        if decision is not None and decision.action == "ready":
+            record = await store.get_session_record(session_id)
+            plans = await store.list_plans(session_id)
+            if not plans and record and not record.get("selected_plan_id"):
+                command = await store.enqueue_plan_request(
+                    session_id,
+                    idempotency_key=f"{key}:autoplan",
+                    expected_revision=decision.requirement.revision,
+                )
+                if command.get("status") == "pending":
+                    await PlanProposalService(
+                        store,
+                        agent_store=getattr(request.app.state, "agent_store", None),
+                        config=get_app_config().report_collaboration,
+                        llm_factory=getattr(request.app.state, "report_collaboration_planner_llm", None),
+                    ).fulfill(session_id, user_id=user_id, command_id=command["command_id"], idempotency_key=f"{key}:autoplan")
         return created
     except Exception as exc:
         return _map_store_error(exc)
@@ -333,6 +355,8 @@ async def request_plans(
             await PlanProposalService(
                 store,
                 agent_store=getattr(request.app.state, "agent_store", None),
+                config=get_app_config().report_collaboration,
+                llm_factory=getattr(request.app.state, "report_collaboration_planner_llm", None),
             ).fulfill(session_id, user_id=user_id, command_id=command["command_id"], idempotency_key=key)
         return Response(status_code=204)
     except Exception as exc:

offline-backend-20260512/backend/app/report_collaboration/agentscope_runtime/model_adapter.py

@@ -91,6 +91,62 @@ class AttemptUsage:
     error_type: str | None = None
     error_message: str | None = None
 
+
+# AgentScope 的 Agent.compress_context 会读 model.context_size 并调 count_tokens。
+_DEFAULT_CONTEXT_SIZE = 128_000
+_CJK_RE = re.compile(r"[\u4e00-\u9fff\u3040-\u30ff\uac00-\ud7af]")
+
+
+def estimate_tokens_for_text(text: str) -> int:
+    """CJK-aware local token estimate: ~1 token per CJK char, ~4 bytes otherwise."""
+
+    if not text:
+        return 0
+    cjk = len(_CJK_RE.findall(text))
+    other = len(text) - cjk
+    return cjk + other // 4 + (1 if other % 4 else 0)
+
+
+def _collect_countable_text(messages: list[Any]) -> str:
+    parts: list[str] = []
+    for msg in messages or []:
+        blocks: Any = None
+        getter = getattr(msg, "get_content_blocks", None)
+        if callable(getter):
+            try:
+                blocks = getter()
+            except Exception:
+                blocks = None
+        if blocks is not None:
+            for block in blocks:
+                if isinstance(block, str):
+                    parts.append(block)
+                elif isinstance(block, dict):
+                    for key in ("text", "thinking", "hint"):
+                        value = block.get(key)
+                        if isinstance(value, str):
+                            parts.append(value)
+                        elif isinstance(value, list):
+                            parts.extend(str(item) for item in value)
+                else:
+                    text = getattr(block, "text", None) or getattr(block, "thinking", None)
+                    if text:
+                        parts.append(str(text))
+            continue
+        if isinstance(msg, dict):
+            parts.append(str(msg.get("content") or ""))
+        else:
+            raw = getattr(msg, "content", "")
+            if isinstance(raw, list):
+                for block in raw:
+                    if isinstance(block, dict):
+                        parts.append(str(block.get("text") or ""))
+                    else:
+                        parts.append(str(getattr(block, "text", block) if not isinstance(block, str) else block))
+            else:
+                parts.append(str(raw or ""))
+    return "\n".join(parts)
+
     def as_dict(self) -> dict[str, Any]:
         return redact_mapping(
             {
@@ -244,6 +300,21 @@ def to_lc_messages(messages: list[Any]) -> list[Any]:
     return converted
 
 
+def _resolve_context_size(app_config: AppConfig | None, config_name: str) -> int:
+    """Best-effort context window from model config extras; sane default otherwise."""
+
+    try:
+        cfg = (app_config or get_app_config()).get_model_config(config_name)
+        extra = cfg.model_dump() if cfg is not None else {}
+        for key in ("context_size", "context_length", "context_window", "max_context_tokens"):
+            value = extra.get(key)
+            if isinstance(value, (int, float)) and int(value) > 0:
+                return int(value)
+    except Exception:
+        pass
+    return _DEFAULT_CONTEXT_SIZE
+
+
 class DeerFlowChatModelAdapter:
     """Frozen DeerFlow-backed chat model. Construct once per node attempt from a snapshot."""
 
@@ -266,6 +337,21 @@ class DeerFlowChatModelAdapter:
         self._llm_factory = llm_factory
         self._node_run_id = node_run_id
         self.last_usage = AttemptUsage(model_name=snapshot.config_name, node_run_id=node_run_id)
+        # AgentScope compress_context 契约：读 context_size 判断是否压缩。
+        self.context_size = _resolve_context_size(app_config, snapshot.config_name)
+
+    async def count_tokens(self, messages: list[Any], tools: list[dict[str, Any]] | None = None) -> int:
+        """AgentScope 契约方法（compress_context 调用）。本地估算，不打 API。"""
+
+        import json as _json
+
+        total = estimate_tokens_for_text(_collect_countable_text(messages))
+        if tools:
+            try:
+                total += estimate_tokens_for_text(_json.dumps(tools, ensure_ascii=False, default=str))
+            except Exception:
+                total += sum(estimate_tokens_for_text(str(item)) for item in tools)
+        return total
 
     def _config(self) -> AppConfig:
         return self._app_config or get_app_config()

offline-backend-20260512/backend/app/report_collaboration/execution/executor.py

@@ -47,7 +47,18 @@ class ReportCollaborationExecutor:
         existing = self._tasks.get(params.run_id)
         if existing is not None and not existing.done():
             return False
-        self._tasks[params.run_id] = asyncio.create_task(self.execute(params), name=f"rc-run-{params.run_id}")
+        task = asyncio.create_task(self.execute(params), name=f"rc-run-{params.run_id}")
+
+        def _log_crash(finished: asyncio.Task[str]) -> None:
+            if not finished.cancelled() and finished.exception() is not None:
+                logger.error(
+                    "rc executor task crashed for run %s",
+                    params.run_id,
+                    exc_info=finished.exception(),
+                )
+
+        task.add_done_callback(_log_crash)
+        self._tasks[params.run_id] = task
         return True
 
     async def drain(self) -> None:

offline-backend-20260512/backend/app/report_collaboration/execution/ledger.py

@@ -10,7 +10,7 @@ from app.report_collaboration.execution.board import ArtifactBoard
 from app.report_collaboration.execution.completion import late_output_may_write_completed, require_can_complete_node
 from app.report_collaboration.execution.quality_gate import QualityGate, QualityGateContext, requirement_from_session_record
 from deerflow.persistence.report_collaboration.base import ReportCollaborationStore
-from deerflow.persistence.report_collaboration.codec import iso_now, make_node_run_id
+from deerflow.persistence.report_collaboration.codec import _TERMINAL_RUN_STATUSES, iso_now, make_node_run_id
 
 TaskResultKind = Literal["artifact_draft", "clarification_request", "repair_request", "failure", "advice"]
 
@@ -314,6 +314,7 @@ class TaskLedger:
         command_id: str,
         instruction: str,
         node_ids: list[str],
+        cascade_downstream: bool = False,
     ) -> list[str]:
         """Supersede affected attempts and persist an immutable command overlay.
 
@@ -321,6 +322,10 @@ class TaskLedger:
         ``superseded`` and is therefore stored only as superseded evidence.
         User-directed rework is not counted against automatic quality-repair
         limits; it creates the next monotonic attempt number.
+
+        ``cascade_downstream=True`` additionally supersedes every transitive
+        dependent of ``node_ids`` and re-plans it, so the wave loop re-runs
+        the whole affected sub-DAG in dependency order (设计 §4.2 指定重跑).
         """
 
         plan, run, latest, artifacts = await self._load(run_id)
@@ -407,8 +412,112 @@ class TaskLedger:
                 active_dedupe_key=str(run["session_id"]),
                 session_status="running",
             )
+        if cascade_downstream and applied:
+            applied.extend(
+                await self._cascade_downstream(
+                    run,
+                    plan,
+                    latest,
+                    artifacts,
+                    roots=[node_id for node_id in node_ids if node_id in applied],
+                    command_id=command_id,
+                )
+            )
         return applied
 
+    async def _cascade_downstream(
+        self,
+        run: dict[str, Any],
+        plan: dict[str, Any],
+        latest: dict[str, dict[str, Any]],
+        artifacts: list[dict[str, Any]],
+        *,
+        roots: list[str],
+        command_id: str,
+    ) -> list[str]:
+        """Supersede transitive dependents of ``roots`` and re-plan them.
+
+        Planned/reopened dependents are left alone — they have produced
+        nothing yet and pick up the fresh upstream artifacts when queued.
+        Like the rework target itself, user-directed cascade re-planning is
+        not budget-limited (monotonic attempt number).
+        """
+
+        deps = _dependency_map(plan)
+        specs = _plan_nodes(plan)
+        dependents: set[str] = set()
+        changed = True
+        while changed:
+            changed = False
+            for node_id, parents in deps.items():
+                if node_id in dependents or node_id in roots:
+                    continue
+                if any(parent in dependents or parent in roots for parent in parents):
+                    dependents.add(node_id)
+                    changed = True
+        cascaded: list[str] = []
+        for node_id in sorted(dependents):
+            node = latest.get(node_id)
+            spec = specs.get(node_id)
+            if node is None or spec is None:
+                continue
+            status = str(node.get("status") or "")
+            if status in {"planned", "reopened", "superseded", "cancelled", "cancel_requested", ""}:
+                continue
+            require_transition(status, "superseded")
+            node["status"] = "superseded"
+            node["ended_at"] = iso_now()
+            node["awaiting_input"] = None
+            await self._store.save_node_run(node)
+            for artifact in artifacts:
+                if (
+                    artifact.get("node_id") == node_id
+                    and int(artifact.get("attempt") or 0) == int(node.get("attempt") or 0)
+                    and not artifact.get("superseded_by")
+                ):
+                    await self._board.mark(artifact, validation_status="superseded", superseded_by=f"command:{command_id}")
+            await self._store.append_event(
+                session_id=str(run["session_id"]),
+                run_id=str(run["id"]),
+                event_type="node.status.changed",
+                data={"node_id": node_id, "node_run_id": node["node_run_id"], "status": "superseded", "reason": "downstream_cascade"},
+            )
+            nxt = int(node.get("attempt") or 0) + 1
+            await self._store.save_node_run(
+                {
+                    "node_run_id": make_node_run_id(str(run["id"]), node_id, nxt),
+                    "session_id": run["session_id"],
+                    "run_id": run["id"],
+                    "node_id": node_id,
+                    "attempt": nxt,
+                    "status": "planned",
+                    "label": node.get("label") or spec.get("label") or node_id,
+                    "role_display_name": node.get("role_display_name") or spec.get("role_display_name"),
+                    "angle": node.get("angle") or spec.get("angle"),
+                    "contract_json": node.get("contract_json"),
+                    "input_artifact_ids": [],
+                    "command_overlay_revision": int(node.get("command_overlay_revision") or 0),
+                    "agent_snapshot": dict(node.get("agent_snapshot") or {}),
+                    "error": None,
+                    "started_at": None,
+                    "ended_at": None,
+                }
+            )
+            await self._store.append_event(
+                session_id=str(run["session_id"]),
+                run_id=str(run["id"]),
+                event_type="node.attempt.created",
+                data={
+                    "node_id": node_id,
+                    "node_run_id": make_node_run_id(str(run["id"]), node_id, nxt),
+                    "attempt": nxt,
+                    "reason": "downstream_cascade",
+                    "command_id": command_id,
+                },
+            )
+            cascaded.append(node_id)
+        return cascaded
+
     @staticmethod
     def _add_command_overlay(node: dict[str, Any], *, command_id: str, instruction: str) -> None:
         snapshot = dict(node.get("agent_snapshot") or {})
@@ -622,8 +731,18 @@ class TaskLedger:
                 return False
         return True
 
+    async def refresh_run_terminal(self, run_id: str) -> str:
+        """Public terminal-state evaluation; safe to call repeatedly."""
+
+        return await self._refresh_run_terminal(run_id)
+
     async def _refresh_run_terminal(self, run_id: str) -> str:
-        _plan, run, latest, _artifacts = await self._load(run_id)
+        plan, run, latest, _artifacts = await self._load(run_id)
+        run_status = str(run.get("status") or "running")
+        if run_status in _TERMINAL_RUN_STATUSES:
+            # Idempotency guard: never re-finalize (the finalizer would run
+            # twice when the executor's wave loop re-checks after completion).
+            return run_status
         statuses = [item["status"] for item in latest.values()]
         if statuses and all(status == "completed" for status in statuses):
             from app.report_collaboration.reporting.finalizer import ReportFinalizationError, ReportFinalizer
@@ -636,11 +755,32 @@ class TaskLedger:
             await self._store.patch_run(run_id, status="completed", active_dedupe_key=None, session_status="completed")
             return "completed"
         if any(status in _BLOCKING or status == "queued" for status in statuses):
-            return str(run.get("status") or "running")
+            return run_status
         if any(status == "failed" for status in statuses):
-            await self._store.patch_run(run_id, status="failed", active_dedupe_key=None, session_status="failed")
+            await self._fail_run(run_id, run, reason="node_failed")
             return "failed"
-        return str(run.get("status") or "running")
+        # Zombie guard: an ``interrupted`` node whose attempt budget is
+        # exhausted can never run again — the ledger only queues
+        # planned/reopened attempts and recovery only opens a new attempt
+        # while attempt < max_attempts. Without this, the run would stay
+        # "running" forever while the dispatcher keeps re-claiming it.
+        specs = _plan_nodes(plan)
+        for node_id, item in latest.items():
+            if str(item.get("status")) != "interrupted":
+                continue
+            if int(item.get("attempt") or 1) >= _node_max_attempts(specs.get(node_id), item):
+                await self._fail_run(run_id, run, reason="node_attempts_exhausted", node_id=node_id)
+                return "failed"
+        return run_status
+
+    async def _fail_run(self, run_id: str, run: dict[str, Any], *, reason: str, node_id: str | None = None) -> None:
+        await self._store.patch_run(run_id, status="failed", active_dedupe_key=None, session_status="failed")
+        await self._store.append_event(
+            session_id=str(run.get("session_id") or ""),
+            run_id=run_id,
+            event_type="run.status.changed",
+            data={"status": "failed", "reason": reason, **({"node_id": node_id} if node_id else {})},
+        )
 
     async def _load(self, run_id: str) -> tuple[dict[str, Any], dict[str, Any], dict[str, dict[str, Any]], list[dict[str, Any]]]:
         run = await self._store.get_run(run_id)
@@ -666,6 +806,20 @@ def _plan_nodes(plan: dict[str, Any]) -> dict[str, dict[str, Any]]:
     return {str(item["id"]): item for item in (plan.get("nodes") or []) if item.get("id")}
 
 
+def _node_max_attempts(spec: dict[str, Any] | None, node_record: dict[str, Any]) -> int:
+    """Same precedence as ``_queue_node``: node contract, then plan spec, then 3."""
+
+    contract = node_record.get("contract_json") or {}
+    if isinstance(contract, str):
+        try:
+            import json as _json
+
+            contract = _json.loads(contract)
+        except Exception:
+            contract = {}
+    return int(contract.get("max_attempts") or (spec or {}).get("max_attempts") or 3)
+
+
 def _dependency_map(plan: dict[str, Any]) -> dict[str, list[str]]:
     deps: dict[str, list[str]] = {node_id: list(spec.get("depends_on") or []) for node_id, spec in _plan_nodes(plan).items()}
     for edge in plan.get("edges") or []:

offline-backend-20260512/backend/app/report_collaboration/execution/member_runner.py

@@ -195,8 +195,11 @@ class WaveExecutor:
                 await self._command_processor.process_safe_point(run_id)
             wave = await self._ledger.ready_assignments(run_id)
             if not wave:
-                run = await self._store.get_run(run_id)
+                # Nothing runnable. Evaluate terminal state: all-completed
-                return str((run or {}).get("status") or "queued")
+                # finalizes, exhausted-interrupted nodes fail the run (zombie
+                # guard). Idempotent when the run already reached a terminal
+                # status via apply_result.
+                return await self._ledger.refresh_run_terminal(run_id)
             for assignment in wave:
                 self.dispatched.append(assignment.node_id)
                 result = await self._runner.run(assignment)

offline-backend-20260512/backend/app/report_collaboration/interventions/service.py

@@ -386,6 +386,7 @@ class RuntimeCommandService:
             command_id=command["command_id"],
             instruction=instruction,
             node_ids=impact_nodes,
+            cascade_downstream=True,
         )
         if not changed:
             raise ReportCollaborationValidationError("COMMAND_NO_EFFECT", "目标节点已过期或命令已经应用")

offline-backend-20260512/backend/app/report_collaboration/planning/planner_agent.py

@@ -0,0 +1,422 @@
+"""规划师智能体（PlannerAgent）——真智能体，不是规则路由 + 模板文案。
+
+设计 v2 的需求对话与工作流生成主路径：
+- ``converse``：与用户多轮自由对话确认需求。回复文案（含开场）完全由
+  智能体生成并直接落库展示 —— 「收到您的报告需求。为了给出…」这类
+  死模板文案已从该路径移除。
+- ``design_workflow``：带着**完整对话上下文**（规划师问了什么、用户
+  答了什么）生成受词表约束的自定义工作流（``StrategyRoleSet``），
+  与确定性模板走同一条 装配 → 校验 → 绑定 链路。
+
+规则路由与模板只作为模型彻底不可用时的最后防线，不出现在用户面前。
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import re
+from typing import Any, Callable
+
+from pydantic import BaseModel, ConfigDict, Field, ValidationError
+
+from app.report_collaboration.agentscope_runtime.model_resolver import ROLE_MODEL_FIELD, RoleModelSnapshot, resolve_role_model
+from app.report_collaboration.contracts.plans import RoleDemand
+from app.report_collaboration.contracts.requirements import ReportRequirementSnapshot
+from app.report_collaboration.planning.controller import RoleSlot, StrategyRoleSet
+from app.report_collaboration.requirement.intent_router import _merge_snapshot, apply_default_assumptions
+from deerflow.config.app_config import AppConfig, get_app_config
+
+logger = logging.getLogger(__name__)
+
+# ── 词表（结构化输出的安全边界，LLM 不受约束的自由只在文案层） ─────────────
+
+MEMBER_ROLES: tuple[str, ...] = tuple(key for key in ROLE_MODEL_FIELD if key != "coordinator")
+
+ROLE_ARTIFACTS: dict[str, tuple[str, ...]] = {
+    "researcher": ("EvidenceBundle",),
+    "verifier": ("VerificationResult",),
+    "analyst": ("AngleAnalysis",),
+    "synthesizer": ("SynthesisOutline",),
+    "writer": ("ReportSectionDraft",),
+    "reviewer": ("ReviewDecision",),
+    "reviser": ("ReportSectionDraft",),
+}
+
+QUALITY_GATES: tuple[str, ...] = ("cite_required", "dual_verification", "independent_review", "coverage_matrix")
+
+# ── 智能体输出契约 ─────────────────────────────────────────────────────────
+
+
+class PlannerTurn(BaseModel):
+    """规划师一轮对话的完整产出：自由文案 + 结构化状态。"""
+
+    model_config = ConfigDict(extra="ignore")
+
+    action: str = "clarify"
+    reply: str = ""
+    questions: list[str] = Field(default_factory=list)
+    impact: list[str] = Field(default_factory=list)
+    assumptions: list[str] = Field(default_factory=list)
+    requirement: dict[str, Any] = Field(default_factory=dict)
+
+
+class StrategyNodeDraft(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    slot_id: str
+    role_key: str
+    display_name: str
+    angle: str | None = None
+    mission: str
+    depends_on: list[str] = Field(default_factory=list)
+    output_artifact_type: str
+    suggested_tools: list[str] = Field(default_factory=list)
+
+
+class StrategyDraft(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    reply: str = ""
+    strategy: str = "custom"
+    title: str
+    summary: str
+    rationale: str
+    estimated_duration_seconds: int = 900
+    estimated_cost_level: str = "medium"
+    quality_gate_ids: list[str] = Field(default_factory=lambda: ["cite_required"])
+    nodes: list[StrategyNodeDraft] = Field(default_factory=list)
+
+
+LlmFactory = Callable[[], Any]
+
+
+def planner_model_snapshot(app_config: AppConfig | None = None) -> RoleModelSnapshot:
+    """规划师身份沿用角色解析链：coordinator_model → planner_model → 默认模型。"""
+
+    return resolve_role_model("coordinator", app_config=app_config or get_app_config())
+
+
+def build_planner_llm(app_config: AppConfig | None = None) -> Any:
+    from deerflow.models.factory import create_chat_model
+
+    snapshot = planner_model_snapshot(app_config)
+    return create_chat_model(snapshot.config_name, app_config=app_config or get_app_config())
+
+
+_THINK_BLOCK_RE = re.compile(r"<think>.*?</think>", re.DOTALL | re.IGNORECASE)
+
+
+def _extract_json(text: str) -> dict[str, Any] | None:
+    if not text:
+        return None
+    # 思考模型（MiniMax-M2.7 等）回复可能带 <think>…</think>；JSON 在思考块之后。
+    # 先剥思考块提取；剥完没有再回退用全文（极端情况 JSON 混在思考内）。
+    for candidate in (_THINK_BLOCK_RE.sub("", text).strip(), text.strip()):
+        if not candidate:
+            continue
+        cleaned = candidate
+        if cleaned.startswith("```"):
+            cleaned = cleaned.strip("`")
+            if cleaned.lower().startswith("json"):
+                cleaned = cleaned[4:]
+            cleaned = cleaned.strip()
+        start = cleaned.find("{")
+        end = cleaned.rfind("}")
+        if start < 0 or end <= start:
+            continue
+        try:
+            parsed = json.loads(cleaned[start : end + 1])
+        except Exception:
+            continue
+        if isinstance(parsed, dict):
+            return parsed
+    return None
+
+
+async def _invoke_json(llm: Any, messages: list[Any]) -> tuple[dict[str, Any] | None, str]:
+    """调用模型并提取 JSON。返回 (解析结果, 原始文本)；解析失败结果为 None。"""
+
+    try:
+        response = await llm.ainvoke(messages)
+        raw = getattr(response, "content", response)
+        if isinstance(raw, list):
+            raw = "".join(str(getattr(block, "text", block) if not isinstance(block, str) else block) for block in raw)
+        raw = str(raw)
+        return _extract_json(raw), raw
+    except Exception as exc:  # noqa: BLE001 — 智能体故障必须显式失败
+        logger.warning("rc planner agent llm call failed: %s: %s", exc.__class__.__name__, exc)
+        return None, ""
+
+
+async def _invoke_json_with_retry(llm: Any, system: str, user_prompt: str) -> dict[str, Any] | None:
+    """带一次格式重试：模型输出不含 JSON 时，附原始输出要求只出 JSON 再问一次。"""
+
+    parsed, raw = await _invoke_json(llm, [("system", system), ("human", user_prompt)])
+    if parsed is None:
+        logger.warning("rc planner reply without JSON (raw head: %.200s)", raw[:200])
+        retry_prompt = (
+            user_prompt
+            + "\n\n【格式提醒】你上一轮的输出没有包含合法 JSON。这次必须只输出一个 JSON 对象，"
+            + f"不要在 JSON 之外输出任何文字。你上一轮的输出是：{raw[:500]}"
+        )
+        parsed, _ = await _invoke_json(llm, [("system", system), ("human", retry_prompt)])
+    return parsed
+
+
+# ── 规划师智能体 ───────────────────────────────────────────────────────────
+
+
+class PlannerAgent:
+    """需求对话与工作流设计的唯一主路径（真智能体）。"""
+
+    def __init__(self, llm: Any, *, app_config: AppConfig | None = None) -> None:
+        self._llm = llm
+        self._app_config = app_config
+
+    # ── 需求对话 ───────────────────────────────────────────────────────────
+
+    async def converse(
+        self,
+        *,
+        user_text: str,
+        history: list[dict[str, str]],
+        previous_snapshot: dict[str, Any] | None,
+        previous_revision: int,
+        clarification_round: int,
+        max_rounds: int,
+    ) -> PlannerTurn | None:
+        """一轮规划对话。返回 None 表示智能体不可用（调用方走最后防线）。"""
+
+        forced = clarification_round >= max_rounds
+        system = _CONVERSE_SYSTEM.format(
+            roles=", ".join(MEMBER_ROLES),
+            forced_note=(
+                "\n\n【重要】澄清轮次已达上限，本轮你必须判定需求就绪（action=ready）："
+                "自然地告诉用户你将按当前理解开始规划，把未能确认的点写进 assumptions。"
+                if forced
+                else ""
+            ),
+        )
+        known = {k: v for k, v in (previous_snapshot or {}).items() if v not in (None, "", [])}
+        known.pop("revision", None)
+        history_text = "\n".join(
+            f"{'用户' if item.get('role') == 'human' else '规划师'}: {str(item.get('content') or '')[:400]}"
+            for item in history[-10:]
+        )
+        user_prompt = (
+            f"当前已确认的需求快照（JSON）：{json.dumps(known, ensure_ascii=False)}\n\n"
+            f"最近的对话：\n{history_text or '（无）'}\n\n"
+            f"用户最新消息：{user_text}"
+        )
+        parsed = await _invoke_json_with_retry(self._llm, system, user_prompt)
+        if parsed is None:
+            return None
+        try:
+            turn = PlannerTurn.model_validate(parsed)
+        except ValidationError:
+            logger.warning("rc planner turn invalid: %s", str(parsed)[:200])
+            return None
+        if not (turn.reply or "").strip():
+            return None
+        if forced and turn.action != "ready":
+            turn.action = "ready"
+        if turn.action == "ready" and not turn.assumptions:
+            turn.assumptions = ["按对话中已确认的信息直接规划"]
+        return turn
+
+    def merge_requirement(
+        self,
+        turn: PlannerTurn,
+        *,
+        previous_snapshot: dict[str, Any] | None,
+        previous_revision: int,
+    ) -> ReportRequirementSnapshot:
+        """把智能体本轮带回的需求字段合并进既有快照（revision 语义同规则路由）。"""
+
+        known = {k: v for k, v in (previous_snapshot or {}).items() if k != "revision"}
+        # LLM 契约是「只填能确认的字段」；再过滤一次空值，避免空串覆盖已确认信息。
+        patch = {
+            k: v
+            for k, v in (turn.requirement or {}).items()
+            if k != "revision" and v not in (None, "", [])
+        }
+        revision = (
+            max(1, int(previous_revision) + 1)
+            if (previous_snapshot or patch)
+            else max(1, int(previous_revision) or 1)
+        )
+        if not previous_snapshot:
+            revision = 1
+        merged = _merge_snapshot(known, patch, revision=revision)
+        if turn.action == "ready":
+            merged = apply_default_assumptions(merged)[0]
+        return merged
+
+    # ── 工作流设计 ─────────────────────────────────────────────────────────
+
+    async def design_workflow(
+        self,
+        *,
+        requirement: ReportRequirementSnapshot,
+        history: list[dict[str, str]],
+        max_team_members: int = 8,
+    ) -> tuple[StrategyRoleSet, str] | None:
+        """带着完整对话上下文设计工作流；返回 (策略, 展示文案) 或 None。"""
+
+        fields = {k: v for k, v in requirement.model_dump().items() if v not in (None, "", [])}
+        fields.pop("revision", None)
+        history_text = "\n".join(
+            f"{'用户' if item.get('role') == 'human' else '规划师'}: {str(item.get('content') or '')[:300]}"
+            for item in history[-12:]
+        )
+        system = _DESIGN_SYSTEM.format(
+            roles=", ".join(MEMBER_ROLES),
+            artifacts="\n".join(f"  - {role}: {' / '.join(arts)}" for role, arts in ROLE_ARTIFACTS.items()),
+            gates=", ".join(QUALITY_GATES),
+            max_nodes=min(max_team_members, 8),
+        )
+        user_prompt = (
+            f"需求快照（JSON）：{json.dumps(fields, ensure_ascii=False)}\n\n"
+            f"需求确认阶段的对话（你与用户）：\n{history_text or '（无）'}\n\n请设计工作流。"
+        )
+        parsed = await _invoke_json_with_retry(self._llm, system, user_prompt)
+        if parsed is None:
+            return None
+        try:
+            draft = StrategyDraft.model_validate(parsed)
+        except ValidationError:
+            logger.warning("rc planner strategy draft invalid: %s", str(parsed)[:200])
+            return None
+        strategy = _validate_strategy_dag(draft)
+        if strategy is None:
+            logger.warning("rc planner strategy failed vocabulary/DAG validation: %s", str(parsed)[:200])
+            return None
+        reply = (draft.reply or "").strip() or f"已按需求设计好协作工作流「{draft.title}」。"
+        return strategy, reply
+
+    def workflow_reply_fallback(self, strategy: StrategyRoleSet) -> str:
+        return (
+            f"我按确认的需求设计了「{strategy.title}」：{strategy.summary}\n"
+            "可以直接采用开始执行；想调整就告诉我（例如「加一条竞品对比线」）。"
+        )
+
+
+# ── 提示词 ─────────────────────────────────────────────────────────────────
+
+_CONVERSE_SYSTEM = """你是「报告协作工作台」的规划师智能体。用户想让你组织一个多智能体团队协作产出一份报告，你的任务是先通过自然对话把需求弄清楚，然后（由你自己）设计协作工作流。
+
+你的对话风格：
+- 像一位资深顾问：先回应用户刚说的话，再追问。回复里**不要**出现「收到您的报告需求」这类机械开场——用你自己的话自然表达。
+- 每轮最多追问 2~3 个真正影响方案设计的问题：关注对象与范围、给谁看、必须覆盖的角度、篇幅与深度、时间范围、特殊约束。用户已说清的绝不重复问。
+- 用户回复后先简短确认你的理解（一两句），再决定是否继续问。
+- 信息足够时，自然地总结你对需求的把握并判定就绪（action=ready），告诉用户下一步你会为他们设计工作流。
+
+可用的团队角色（供你构思，不必对用户罗列）：{roles}{forced_note}
+
+每轮严格输出一个 JSON 对象（不要输出 JSON 以外的文字）：
+{{
+  "reply": "给用户看的完整回复（自然对话，直接展示）",
+  "action": "clarify" 或 "ready",
+  "questions": ["追问的问题（与 reply 一致，供系统记录）"],
+  "impact": ["为什么需要问（与 questions 对应）"],
+  "assumptions": ["ready 时：你采用的默认假设"],
+  "requirement": {{"topic": "...", "objective": "...", "audience": "...", "required_angles": ["..."], "time_range": "...", "length_target": "...", "tone": "..."}}
+}}
+
+requirement 只填你从对话中能确认的字段（含此前轮次已确认的），其余省略。reply 用中文。"""
+
+_DESIGN_SYSTEM = """你是「报告协作工作台」的规划师智能体。需求已经与你确认完毕，现在请设计多智能体协作工作流。
+
+词表约束（结构化部分必须严格遵守）：
+- role_key 只能取：{roles}
+- 各角色的 output_artifact_type 只能取：
+{artifacts}
+- quality_gate_ids 只能取：{gates}
+
+结构要求：
+- 2~{max_nodes} 个节点；至少 1 个 writer；正式报告建议加 reviewer。
+- 研究线（researcher）按角度并行；research → verify → analyze → write → review 是常见主干，可按需求裁剪。
+- depends_on 引用其他节点的 slot_id；禁止环；至少一个节点 depends_on 为空（起点）。
+- slot_id 用简短英文（如 research-market / write-main）。mission 写清楚该节点做什么、产出什么。
+
+严格输出 JSON（不要输出 JSON 以外的文字）：
+{{
+  "reply": "给用户看的介绍：说明你为什么这样设计（对应需求的哪些要求），一两句即可",
+  "strategy": "custom",
+  "title": "方案标题",
+  "summary": "一句话概述团队构成与链路",
+  "rationale": "设计理由",
+  "estimated_duration_seconds": 1500,
+  "estimated_cost_level": "low|medium|high",
+  "quality_gate_ids": ["cite_required"],
+  "nodes": [
+    {{"slot_id": "research-market", "role_key": "researcher", "display_name": "研究员", "angle": "市场份额", "mission": "...", "depends_on": [], "output_artifact_type": "EvidenceBundle", "suggested_tools": ["web_search"]}}
+  ]
+}}"""
+
+
+# ── 工作流草稿校验（词表 + DAG） ──────────────────────────────────────────
+
+
+def _validate_strategy_dag(draft: StrategyDraft) -> StrategyRoleSet | None:
+    nodes = draft.nodes
+    if not 2 <= len(nodes) <= 8:
+        return None
+    slot_ids = {node.slot_id for node in nodes}
+    if len(slot_ids) != len(nodes):
+        return None
+    if not any(node.role_key == "writer" for node in nodes):
+        return None
+    for node in nodes:
+        if node.role_key not in MEMBER_ROLES:
+            return None
+        if node.output_artifact_type not in ROLE_ARTIFACTS.get(node.role_key, ()):
+            return None
+        for dep in node.depends_on:
+            if dep not in slot_ids or dep == node.slot_id:
+                return None
+    indegree = {node.slot_id: len(node.depends_on) for node in nodes}
+    children: dict[str, list[str]] = {node.slot_id: [] for node in nodes}
+    for node in nodes:
+        for dep in node.depends_on:
+            children[dep].append(node.slot_id)
+    queue = [slot for slot, degree in indegree.items() if degree == 0]
+    seen = 0
+    while queue:
+        current = queue.pop()
+        seen += 1
+        for child in children[current]:
+            indegree[child] -= 1
+            if indegree[child] == 0:
+                queue.append(child)
+    if seen != len(nodes):
+        return None
+
+    gates = [gate for gate in draft.quality_gate_ids if gate in QUALITY_GATES] or ["cite_required"]
+    return StrategyRoleSet(
+        strategy="custom" if draft.strategy == "custom" else f"custom_{draft.strategy[:24]}",
+        title=draft.title[:80],
+        summary=draft.summary[:200],
+        rationale=draft.rationale[:300],
+        estimated_duration_seconds=max(300, min(int(draft.estimated_duration_seconds or 1500), 7200)),
+        estimated_cost_level=draft.estimated_cost_level if draft.estimated_cost_level in {"low", "medium", "high"} else "medium",
+        recommended=True,
+        slots=[
+            RoleSlot(
+                slot_id=node.slot_id,
+                demand=RoleDemand(
+                    role_key=node.role_key,
+                    display_name=node.display_name[:32] or node.role_key,
+                    angle=(node.angle or None),
+                    mission=node.mission[:400],
+                    depends_on_roles=[],
+                    output_artifact_type=node.output_artifact_type,
+                    suggested_tools=[tool for tool in node.suggested_tools if tool][:4],
+                ),
+                depends_on=list(node.depends_on),
+            )
+            for node in nodes
+        ],
+        quality_gate_ids=gates,
+    )

offline-backend-20260512/backend/app/report_collaboration/planning/service.py

@@ -1,17 +1,18 @@
 """Fulfill a durable plan_request: catalog → select → assemble → validate → persist.
 
-Does not start a run. Rule-based until the BE-7 model adapter is wired.
+Does not start a run. P1.2 起优先由规划师 LLM 生成定制策略，模板为回退。
 """
 
 from __future__ import annotations
 
+import logging
 from typing import Any
 from uuid import uuid4
 
 from app.report_collaboration.contracts.requirements import ReportRequirementSnapshot
 from app.report_collaboration.planning.assembler import PlanAssembler
 from app.report_collaboration.planning.catalog import project_agent_catalog
-from app.report_collaboration.planning.controller import PlanController
+from app.report_collaboration.planning.planner_agent import LlmFactory
 from app.report_collaboration.planning.validator import apply_validation, validate_assembled_plan
 from app.report_collaboration.requirement.defaults import COORDINATOR_NAME, DEFAULT_TITLE
 from deerflow.config.app_config import get_app_config
@@ -22,6 +23,8 @@ from deerflow.persistence.report_collaboration import (
     ReportCollaborationValidationError,
 )
 
+logger = logging.getLogger(__name__)
+
 _BUSY = frozenset({"running", "awaiting_input", "reviewing", "completed", "failed", "cancelled"})
 
 
@@ -51,12 +54,15 @@ class PlanProposalService:
         *,
         agent_store: Any | None = None,
         config: ReportCollaborationConfig | None = None,
+        llm_factory: LlmFactory | None = None,
     ) -> None:
         self._store = store
         self._agent_store = agent_store
         self._config = config
-        self._controller = PlanController()
         self._assembler = PlanAssembler()
+        self._llm_factory = llm_factory
+        self._llm: Any | None = None
+        self._llm_tried = False
 
     def _cfg(self) -> ReportCollaborationConfig:
         return self._config or get_app_config().report_collaboration
@@ -110,12 +116,29 @@ class PlanProposalService:
         requirement = requirement_from_session(record)
         cfg = self._cfg()
         catalog = await project_agent_catalog(user_id=user_id, config=cfg, agent_store=self._agent_store)
-        strategies = self._controller.propose(
+        group_id = f"rpg_{uuid4().hex[:12]}"
-            requirement,
+
-            max_parallel_tasks=cfg.max_parallel_tasks,
+        # P1-R：规划师智能体是唯一路径 —— 带着完整对话上下文设计工作流。
+        # 模型不可用即无法继续（无模板兜底）。
+        from app.report_collaboration.planning.planner_agent import PlannerAgent
+
+        llm = self._planner_llm()
+        history = [
+            {"role": "human" if item.get("role") == "human" else "ai", "content": str(item.get("content") or "")}
+            for item in await self._store.list_messages(session_id)
+        ]
+        designed = await PlannerAgent(llm).design_workflow(
+            requirement=requirement,
+            history=history,
             max_team_members=cfg.max_team_members,
         )
-        group_id = f"rpg_{uuid4().hex[:12]}"
+        if designed is None:
+            raise ReportCollaborationValidationError(
+                "PLANNER_UNAVAILABLE", "规划师智能体未能生成有效工作流，请重试或换个说法再试"
+            )
+        strategies: list[Any] = [designed[0]]
+        planner_reply: str | None = designed[1]
+
         assembled_list = [
             self._assembler.assemble(
                 strategy,
@@ -140,13 +163,44 @@ class PlanProposalService:
         for candidate in persisted:
             saved.append(await self._store.insert_plan(session_id, candidate.model_dump()))
         await self._store.update_command_status(command_id, status="completed")
+        content = planner_reply or "工作流已生成，可开始执行。"
         await self._store.create_message(
             session_id,
             role="ai",
-            content="已根据当前需求生成 3 个候选协作方案。请选择其中一个后再开始执行；选择方案不会自动开跑。",
+            content=content,
             idempotency_key=f"{idempotency_key}:plans",
             name=COORDINATOR_NAME,
-            metadata={"plan_request_command_id": command_id, "proposal_group_id": group_id},
+            metadata={"plan_request_command_id": command_id, "proposal_group_id": group_id, "plan_source": "planner_agent"},
             completed=True,
         )
         return saved
+
+    def _planner_llm(self) -> Any:
+        """规划师模型：注入优先（测试）；否则按 planner_llm_enabled 懒构建。不可用即抛错。"""
+
+        if self._llm_factory is not None:
+            llm = self._llm_factory()
+            if llm is None:
+                raise ReportCollaborationValidationError("PLANNER_UNAVAILABLE", "规划师智能体不可用，请检查模型配置")
+            return llm
+        try:
+            if not (self._config or get_app_config().report_collaboration).planner_llm_enabled:
+                raise ReportCollaborationValidationError(
+                    "PLANNER_UNAVAILABLE", "规划师智能体未启用（report_collaboration.planner_llm_enabled）"
+                )
+        except ReportCollaborationValidationError:
+            raise
+        except Exception as exc:
+            raise ReportCollaborationValidationError("PLANNER_UNAVAILABLE", f"规划师智能体配置不可读：{exc}") from None
+        if not self._llm_tried:
+            self._llm_tried = True
+            try:
+                from app.report_collaboration.planning.planner_agent import build_planner_llm
+
+                self._llm = build_planner_llm()
+            except Exception as exc:
+                logger.warning("rc planner agent llm build failed: %s", exc)
+                self._llm = None
+        if self._llm is None:
+            raise ReportCollaborationValidationError("PLANNER_UNAVAILABLE", "规划师智能体不可用，请检查模型配置")
+        return self._llm

offline-backend-20260512/backend/app/report_collaboration/requirement/__init__.py

@@ -1,6 +1,6 @@
 """Pre-plan requirement resolution package."""
 
-from app.report_collaboration.requirement.intent_router import PrePlanDecision, route_pre_plan_turn
+from app.report_collaboration.requirement.intent_router import PrePlanDecision
 from app.report_collaboration.requirement.resolver import RequirementResolver
 
-__all__ = ["PrePlanDecision", "RequirementResolver", "route_pre_plan_turn"]
+__all__ = ["PrePlanDecision", "RequirementResolver"]

offline-backend-20260512/backend/app/report_collaboration/requirement/intent_router.py

@@ -1,4 +1,9 @@
-"""Pre-plan IntentRouterAgent: understand the user turn, never dispatch work."""
+"""Pre-plan 数据结构与快照合并工具。
+
+P1-R 之后需求确认完全由规划师智能体（``planning/planner_agent.py``）驱动；
+本模块只保留被复用的数据结构与合并工具。原规则路由 ``route_pre_plan_turn``
+已按设计 v2 移除 —— 模型不可用即报错，不再有规则兜底。
+"""
 
 from __future__ import annotations
 
@@ -7,12 +12,11 @@ from typing import Any, Literal
 from pydantic import BaseModel, ConfigDict, Field
 
 from app.report_collaboration.contracts.requirements import ReportRequirementSnapshot
-from app.report_collaboration.requirement.defaults import DEFAULT_ANGLES, DEFAULT_AUDIENCE, DEFAULT_LANGUAGE, DEFAULT_SOURCE_CONSTRAINTS, DEFAULT_TONE, PLAN_AFFECTING_GAPS
+from app.report_collaboration.requirement.defaults import DEFAULT_ANGLES, DEFAULT_AUDIENCE, DEFAULT_LANGUAGE, DEFAULT_SOURCE_CONSTRAINTS, DEFAULT_TONE
-from app.report_collaboration.requirement.extractor import extract_requirement_patch, wants_direct_start
 
 
 class PrePlanDecision(BaseModel):
-    """Structured output of the pre-plan router. It never builds a graph or starts a run."""
+    """规划师智能体一轮决策。It never builds a graph or starts a run."""
 
     model_config = ConfigDict(extra="forbid")
 
@@ -24,11 +28,13 @@ class PrePlanDecision(BaseModel):
     used_defaults: bool = False
     reason: str
     confidence: float = Field(ge=0, le=1)
+    #: 规划师智能体产出的完整对话文案，直接落库展示。
+    reply: str | None = None
 
 
 def _merge_snapshot(previous: dict[str, Any] | None, patch: dict[str, Any], *, revision: int) -> ReportRequirementSnapshot:
     base = dict(previous or {})
-    base.update(patch)
+    base.update(patch or {})
     topic = str(base.get("topic") or "").strip()
     return ReportRequirementSnapshot(
         topic=topic or "未指定主题",
@@ -49,21 +55,6 @@ def _merge_snapshot(previous: dict[str, Any] | None, patch: dict[str, Any], *, r
     )
 
 
-def _gaps(requirement: ReportRequirementSnapshot) -> list[tuple[str, str, str]]:
-    missing: list[tuple[str, str, str]] = []
-    if not requirement.topic or requirement.topic == "未指定主题":
-        missing.append(("topic", "这份报告要写什么主题？请用一两句话说明对象、时间和目的。", "没有主题无法选择方案"))
-    for field, question, impact in PLAN_AFFECTING_GAPS:
-        value = getattr(requirement, field, None)
-        if field == "required_angles":
-            empty = not value
-        else:
-            empty = not (value or "").strip()
-        if empty:
-            missing.append((field, question, impact))
-    return missing
-
-
 def apply_default_assumptions(requirement: ReportRequirementSnapshot) -> tuple[ReportRequirementSnapshot, list[str]]:
     assumptions: list[str] = []
     data = requirement.model_dump()
@@ -82,62 +73,3 @@ def apply_default_assumptions(requirement: ReportRequirementSnapshot) -> tuple[R
         data["tone"] = DEFAULT_TONE
         assumptions.append(DEFAULT_TONE)
     return ReportRequirementSnapshot.model_validate(data), assumptions
-
-
-def route_pre_plan_turn(
-    text: str,
-    *,
-    previous_snapshot: dict[str, Any] | None,
-    previous_revision: int,
-    clarification_round: int,
-    max_clarification_rounds: int,
-) -> PrePlanDecision:
-    """Rule-based pre-plan routing. Does not call tools or create plans."""
-
-    patch = extract_requirement_patch(text, previous=previous_snapshot)
-    revision = max(1, int(previous_revision) + 1) if previous_snapshot or patch else max(1, int(previous_revision) or 1)
-    if not previous_snapshot:
-        revision = 1
-    requirement = _merge_snapshot(previous_snapshot, patch, revision=revision)
-    skip = wants_direct_start(text)
-    hit_cap = clarification_round >= max_clarification_rounds
-    gaps = _gaps(requirement)
-
-    if requirement.topic == "未指定主题" and skip:
-        return PrePlanDecision(
-            action="clarify",
-            requirement=requirement,
-            questions=["这份报告要写什么主题？请用一两句话说明对象、时间和目的。"],
-            impact=["没有主题无法选择方案"],
-            reason="用户要求直接开始，但还没有主题",
-            confidence=0.9,
-        )
-
-    if skip or (hit_cap and gaps):
-        filled, assumptions = apply_default_assumptions(requirement)
-        return PrePlanDecision(
-            action="ready",
-            requirement=filled,
-            assumptions=assumptions,
-            used_defaults=True,
-            reason="用户要求按默认假设继续" if skip else "已达澄清上限，使用明确假设继续",
-            confidence=0.86 if skip else 0.7,
-        )
-
-    if gaps:
-        return PrePlanDecision(
-            action="clarify",
-            requirement=requirement,
-            questions=[item[1] for item in gaps],
-            impact=[item[2] for item in gaps],
-            reason="缺少真正影响方案选择的信息",
-            confidence=0.8,
-        )
-
-    return PrePlanDecision(
-        action="ready",
-        requirement=requirement,
-        reason="必要信息已齐，可以进入规划",
-        confidence=0.84,
-        used_defaults=False,
-    )

offline-backend-20260512/backend/app/report_collaboration/requirement/resolver.py

@@ -1,18 +1,39 @@
-"""Validate and persist requirement revisions. Never starts planning or a run."""
+"""Requirement resolver: the planner agent is the only path.
+
+模型不可用即无法继续 —— 直接抛 ``PLANNER_UNAVAILABLE``，没有任何规则/
+模板兜底（设计 v2：需求确认必须由真智能体完成）。
+"""
 
 from __future__ import annotations
 
+import logging
 from typing import Any
 from uuid import uuid4
 
 from app.report_collaboration.contracts.clarifications import ClarificationRecord
-from app.report_collaboration.contracts.requirements import ReportRequirementSnapshot
 from app.report_collaboration.requirement.defaults import COORDINATOR_NAME, DEFAULT_TITLE
-from app.report_collaboration.requirement.intent_router import PrePlanDecision, route_pre_plan_turn
+from app.report_collaboration.requirement.intent_router import PrePlanDecision
 from deerflow.config.app_config import get_app_config
-from deerflow.persistence.report_collaboration import ReportCollaborationStore
+from deerflow.persistence.report_collaboration import (
+    ReportCollaborationStore,
+    ReportCollaborationValidationError,
+)
 from deerflow.persistence.report_collaboration.codec import iso_now
 
+logger = logging.getLogger(__name__)
+
+# 仅供类型注解；运行时惰性导入避免 requirement ↔ planning 循环依赖。
+LlmFactory = Any
+
+
+def _title_from_topic(topic: str, current: str | None) -> str | None:
+    if current and current != DEFAULT_TITLE:
+        return None
+    cleaned = (topic or "").strip()
+    if not cleaned or cleaned == "未指定主题":
+        return None
+    return cleaned[:40]
+
 
 def _state_from_session(record: dict[str, Any]) -> dict[str, Any]:
     raw = record.get("requirement_json")
@@ -35,58 +56,56 @@ def _state_from_session(record: dict[str, Any]) -> dict[str, Any]:
     return {"snapshot": None, "clarifications": [], "assumptions": [], "clarification_round": 0, "ready_to_plan": False}
 
 
-def _dump_state(
-    *,
-    snapshot: ReportRequirementSnapshot,
-    clarifications: list[dict[str, Any]],
-    assumptions: list[str],
-    clarification_round: int,
-    ready_to_plan: bool,
-) -> dict[str, Any]:
-    return {
-        "snapshot": snapshot.model_dump(),
-        "clarifications": clarifications,
-        "assumptions": assumptions,
-        "clarification_round": clarification_round,
-        "ready_to_plan": ready_to_plan,
-    }
-
-
-def format_clarification_message(questions: list[str], *, allow_skip: bool = True) -> str:
-    numbered = "\n".join(f"{index}. {question}" for index, question in enumerate(questions, start=1))
-    suffix = "\n\n也可以直接回复「直接开始」，我会按默认假设记录需求。澄清完成不会自动生成方案。" if allow_skip else ""
-    return f"收到您的报告需求。为了给出最合适的协作方案，请先确认：\n\n{numbered}{suffix}"
-
-
-def format_ready_message(decision: PrePlanDecision) -> str:
-    req = decision.requirement
-    angles = "、".join(req.required_angles) if req.required_angles else "（未指定）"
-    lines = [
-        f"已写入需求版本 v{req.revision}。",
-        f"主题：{req.topic}",
-        f"读者：{req.audience or '未指定'}",
-        f"角度：{angles}",
-    ]
-    if decision.assumptions:
-        lines.append("默认假设：" + "；".join(decision.assumptions) + "。")
-    lines.append("澄清完成不等于已经生成候选方案，需要显式发起规划后才会出现方案卡。")
-    return "\n".join(lines)
-
-
-def _title_from_topic(topic: str, current: str | None) -> str | None:
-    if current and current != DEFAULT_TITLE:
-        return None
-    cleaned = (topic or "").strip()
-    if not cleaned or cleaned == "未指定主题":
-        return None
-    return cleaned[:40]
-
-
 class RequirementResolver:
     """Writes requirement revisions and clarification records. Does not enqueue plans."""
 
-    def __init__(self, store: ReportCollaborationStore) -> None:
+    def __init__(
+        self,
+        store: ReportCollaborationStore,
+        *,
+        planner_llm_factory: LlmFactory | None = None,
+        config: Any | None = None,
+    ) -> None:
         self._store = store
+        self._planner_llm_factory = planner_llm_factory
+        self._config = config
+        self._default_llm: Any = None
+        self._default_llm_tried = False
+
+    def _planner_llm(self) -> Any:
+        """规划师模型：注入优先（测试）；否则按 planner_llm_enabled 懒构建。不可用即抛错。"""
+
+        if self._planner_llm_factory is not None:
+            llm = self._planner_llm_factory()
+            if llm is None:
+                raise ReportCollaborationValidationError("PLANNER_UNAVAILABLE", "规划师智能体不可用，请检查模型配置")
+            return llm
+        try:
+            from deerflow.config.report_collaboration_config import ReportCollaborationConfig
+
+            cfg: Any = self._config
+            if cfg is None:
+                cfg = get_app_config().report_collaboration
+            if not ReportCollaborationConfig.model_validate(cfg).planner_llm_enabled:
+                raise ReportCollaborationValidationError(
+                    "PLANNER_UNAVAILABLE", "规划师智能体未启用（report_collaboration.planner_llm_enabled）"
+                )
+        except ReportCollaborationValidationError:
+            raise
+        except Exception as exc:
+            raise ReportCollaborationValidationError("PLANNER_UNAVAILABLE", f"规划师智能体配置不可读：{exc}") from None
+        if not self._default_llm_tried:
+            self._default_llm_tried = True
+            try:
+                from app.report_collaboration.planning.planner_agent import build_planner_llm
+
+                self._default_llm = build_planner_llm()
+            except Exception as exc:
+                logger.warning("rc planner agent llm build failed: %s", exc)
+                self._default_llm = None
+        if self._default_llm is None:
+            raise ReportCollaborationValidationError("PLANNER_UNAVAILABLE", "规划师智能体不可用，请检查模型配置")
+        return self._default_llm
 
     async def handle_user_message(
         self,
@@ -111,16 +130,67 @@ class RequirementResolver:
         cfg_rounds = max_clarification_rounds
         if cfg_rounds is None:
             cfg_rounds = int(get_app_config().report_collaboration.max_clarification_rounds)
-        decision = route_pre_plan_turn(
+
-            str(user_message.get("content") or ""),
+        text = str(user_message.get("content") or "")
-            previous_snapshot=state["snapshot"],
+        history = [
+            {"role": "human" if item.get("role") == "human" else "ai", "content": str(item.get("content") or "")}
+            for item in messages
+            if item.get("message_id") != user_message["message_id"]
+        ]
+
+        decision = await self._agent_decision(
+            text,
+            history=history,
+            state=state,
             previous_revision=int(record.get("requirement_revision") or 0),
-            clarification_round=int(state["clarification_round"] or 0),
+            max_rounds=cfg_rounds,
-            max_clarification_rounds=cfg_rounds,
         )
+        if decision is None:
+            # 模型调用失败/输出不合法 —— 无法继续，直接报错（无兜底）。
+            raise ReportCollaborationValidationError("PLANNER_UNAVAILABLE", "规划师智能体本轮未返回有效决策，请重试")
         await self._apply(session_id, record, state, decision, user_message=user_message, idempotency_key=idempotency_key)
         return decision
 
+    async def _agent_decision(
+        self,
+        text: str,
+        *,
+        history: list[dict[str, str]],
+        state: dict[str, Any],
+        previous_revision: int,
+        max_rounds: int,
+    ) -> PrePlanDecision | None:
+        llm = self._planner_llm()
+        from app.report_collaboration.planning.planner_agent import PlannerAgent
+
+        agent = PlannerAgent(llm)
+        turn = await agent.converse(
+            user_text=text,
+            history=history,
+            previous_snapshot=state["snapshot"],
+            previous_revision=previous_revision,
+            clarification_round=int(state["clarification_round"] or 0),
+            max_rounds=max_rounds,
+        )
+        if turn is None:
+            return None
+        requirement = agent.merge_requirement(
+            turn,
+            previous_snapshot=state["snapshot"],
+            previous_revision=previous_revision,
+        )
+        return PrePlanDecision(
+            action="ready" if turn.action == "ready" else "clarify",
+            requirement=requirement,
+            questions=[item for item in turn.questions if item and item.strip()][:3],
+            impact=[item for item in turn.impact if item][:3],
+            assumptions=[item for item in turn.assumptions if item],
+            used_defaults=bool(turn.assumptions),
+            reason="planner_agent",
+            confidence=0.9,
+            reply=turn.reply.strip(),
+        )
+
     async def _apply(
         self,
         session_id: str,
@@ -146,7 +216,7 @@ class RequirementResolver:
                 created_at=iso_now(),
             )
             clarifications.append(clarification.model_dump())
-            content = format_clarification_message(decision.questions)
+            content = decision.reply or ""
             metadata = {
                 "preplan_for": user_message["message_id"],
                 "clarification": clarification.model_dump(),
@@ -155,17 +225,22 @@ class RequirementResolver:
         else:
             for item in clarifications:
                 item["resolved"] = True
-            content = format_ready_message(decision)
+            content = decision.reply or ""
-            metadata = {"preplan_for": user_message["message_id"], "ready_to_plan": True}
+            metadata = {
+                "preplan_for": user_message["message_id"],
+                "ready_to_plan": True,
+                # 结构化需求快照：前端据此渲染「需求摘要卡」（计划模式 UI）。
+                "requirement": decision.requirement.model_dump(),
+            }
             ready = True
 
-        payload = _dump_state(
+        payload = {
-            snapshot=decision.requirement,
+            "snapshot": decision.requirement.model_dump(),
-            clarifications=clarifications,
+            "clarifications": clarifications,
-            assumptions=list(decision.assumptions),
+            "assumptions": list(decision.assumptions),
-            clarification_round=round_count,
+            "clarification_round": round_count,
-            ready_to_plan=ready,
+            "ready_to_plan": ready,
-        )
+        }
         await self._store.save_requirement_state(
             session_id,
             requirement_json=payload,

offline-backend-20260512/backend/config.yaml

@@ -456,7 +456,10 @@ auth_login:
 # 点击后在新标签页打开下面的 url。enabled 关掉则不显示该入口。
 
 database:
-  backend: sqlite
+  # 本地业务数据使用远程 MySQL；连接地址由 backend/.env.local 提供。
+  # LangGraph 对话检查点仍由下方 checkpointer 保存到本机 SQLite。
+  backend: mysql
+  mysql_url: $MYSQL_DATABASE_URL
   sqlite_dir: .deer-flow/data
   # 高斯/OpenGauss 业务库示例（启用前先执行 uv sync --extra gauss）：
   # backend: gauss
@@ -605,7 +608,7 @@ llmwiki:
     enabled: true
     # 免登录 POST /api/knowledge/vector-search 返回的 Wiki 一键跳转链接前缀。
     frontend_base_url: "http://127.0.0.1:5174"
-    auto_sync: true
+    auto_sync: false
     sync_interval_seconds: 30
     database_batch_size: 50
     embedding:
@@ -748,6 +751,7 @@ workflows:
 report_collaboration:
   enabled: true
   worker_enabled: true
+  planner_llm_enabled: true
   planner_model: null
   coordinator_model: null
   research_model: null

offline-backend-20260512/backend/knowledge/knowledge_generate/.gitignore

@@ -0,0 +1,5 @@
+*.log
+run_summary*.json
+runs/**
+!runs/
+!runs/placeholder_audit.md

offline-backend-20260512/backend/offline-wheels/README-离线安装-asyncmy.md

@@ -1,48 +0,0 @@
-# asyncmy 离线安装说明
-
-本目录用于给无法联网的后端服务器补装 MySQL 异步驱动 `asyncmy==0.2.11`。
-
-项目的 `pyproject.toml` 已经声明该依赖；服务器启动报 `MySQL async driver is not installed`，说明实际运行后端的 Python 虚拟环境尚未安装它。
-
-## 文件选择
-
-- `asyncmy-0.2.11-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl`：Ubuntu、CentOS、RHEL、Debian 等常见 glibc Linux（x86_64）。
-- `asyncmy-0.2.11-cp312-cp312-musllinux_1_2_x86_64.whl`：Alpine Linux（x86_64）。
-
-这两个 wheel 都只适用于 **CPython 3.12、x86_64**。在服务器执行 `python -V` 和 `uname -m` 确认：应分别显示 Python 3.12 和 `x86_64`/`amd64`。
-
-## 安装步骤
-
-1. 将整个 `offline-wheels` 目录复制到后端目录（与 `pyproject.toml` 同级）。
-2. 进入后端实际运行的虚拟环境；若项目使用默认虚拟环境：
-
-   ```bash
-   cd /path/to/backend
-   source .venv/bin/activate
-   ```
-
-3. 普通 Linux（Ubuntu/CentOS/RHEL/Debian）执行：
-
-   ```bash
-   python -m pip install --no-index ./offline-wheels/asyncmy/asyncmy-0.2.11-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-   ```
-
-   Alpine Linux 则执行：
-
-   ```bash
-   python -m pip install --no-index ./offline-wheels/asyncmy/asyncmy-0.2.11-cp312-cp312-musllinux_1_2_x86_64.whl
-   ```
-
-4. 验证：
-
-   ```bash
-   python -c "import asyncmy; print(asyncmy.__version__)"
-   ```
-
-5. 重启后端服务。
-
-## 注意事项
-
-- 必须使用**启动后端服务的同一个 Python/虚拟环境**安装，系统 Python 安装成功并不代表服务能找到该包。
-- 不要使用 `--force-reinstall`；安装缺失的 `asyncmy` 即可。
-- 若服务器是 Python 3.13、ARM/aarch64，或不是上述 Linux 类型，请不要安装本目录 wheel，需要下载对应平台版本。

offline-backend-20260512/backend/packages/harness/deerflow/config/report_collaboration_config.py

@@ -18,6 +18,10 @@ class ReportCollaborationConfig(BaseModel):
         default=False,
         description="When true, a dispatcher/worker may claim run leases. Gateway-only processes keep this false.",
     )
+    planner_llm_enabled: bool = Field(
+        default=False,
+        description="P1: LLM planner conversation (clarify + custom strategy). Off keeps the deterministic rule router and template plans.",
+    )
     planner_model: str | None = Field(default=None, description="Model name from AppConfig.models for planning")
     coordinator_model: str | None = Field(default=None, description="Model name from AppConfig.models for the constrained coordinator")
     research_model: str | None = Field(default=None, description="Model name from AppConfig.models for research members")

offline-backend-20260512/backend/packages/harness/deerflow/persistence/migrations/versions/20260909_01_merge_assistant_wiki_and_report_collaboration_heads.py

@@ -0,0 +1,21 @@
+"""Merge assistant-wiki metadata and report-collaboration heads.
+
+Revision ID: 20260909_01
+Revises: 20260903_01, 20260907_01
+Create Date: 2026-09-09
+"""
+
+from __future__ import annotations
+
+revision = "20260909_01"
+down_revision = ("20260903_01", "20260907_01")
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    pass
+
+
+def downgrade() -> None:
+    pass

offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/__init__.py

@@ -0,0 +1,13 @@
+"""Persistence layer for custom skills (per-user ownership + publish state)."""
+
+from deerflow.persistence.skills.base import SkillStore
+from deerflow.persistence.skills.memory import MemorySkillStore
+from deerflow.persistence.skills.sql import SkillRepository
+
+__all__ = ["SkillStore", "MemorySkillStore", "SkillRepository", "make_skill_store"]
+
+
+def make_skill_store(session_factory):
+    if session_factory is None:
+        return MemorySkillStore()
+    return SkillRepository(session_factory)

offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/base.py

@@ -0,0 +1,107 @@
+"""Abstract persistence API for custom skill ownership + publish state."""
+
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+from typing import Any
+
+
+class SkillStore(ABC):
+    @abstractmethod
+    async def list_visible(self, user_id: str) -> list[dict[str, Any]]:
+        """List all skill records visible to ``user_id``.
+
+        Visible = owner, OR published by anyone, OR built-in (owner_user_id IS NULL).
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def get_visible(self, name: str, user_id: str) -> dict[str, Any] | None:
+        raise NotImplementedError
+
+    @abstractmethod
+    async def get_any(self, name: str) -> dict[str, Any] | None:
+        """Lookup by name regardless of ownership/visibility (admin / cascade use)."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def get_owned(self, name: str, user_id: str) -> dict[str, Any] | None:
+        raise NotImplementedError
+
+    @abstractmethod
+    async def create(self, data: dict[str, Any]) -> dict[str, Any]:
+        """Create a skill record. Caller is responsible for global-name uniqueness check."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def ensure_legacy(self, data: dict[str, Any]) -> dict[str, Any]:
+        """Idempotently insert a legacy on-disk skill into the DB. Updates the row if it
+        already exists with ``owner_user_id IS NULL`` (so a re-run can migrate ownership).
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def update(self, name: str, user_id: str, data: dict[str, Any]) -> dict[str, Any] | None:
+        """Owner-only update. Returns None if the caller doesn't own the skill."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def update_admin(self, name: str, data: dict[str, Any]) -> dict[str, Any] | None:
+        """Admin override — caller is responsible for auth check.
+
+        Unlike ``update_builtin`` for agents, this accepts ANY skill including
+        user-owned ones, since the spec lets admin take down others' skills.
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def delete(self, name: str, user_id: str) -> bool:
+        raise NotImplementedError
+
+    @abstractmethod
+    async def delete_admin(self, name: str) -> bool:
+        """Admin-side delete; bypasses ownership check (caller authorizes)."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def count_owned(self, user_id: str) -> int:
+        """Return the number of skills owned by ``user_id`` (owner_user_id IS NOT NULL)."""
+        raise NotImplementedError
+
+    @abstractmethod
+    async def list_all(self) -> list[dict[str, Any]]:
+        """Return every skill record regardless of ownership/visibility.
+
+        Caller is responsible for authorization (admin only).
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def list_published_names(self) -> set[str]:
+        """Return the set of skill names with ``published = True`` (any owner).
+
+        Batch alternative to calling ``get_any`` per skill — the curator uses
+        this to filter out published skills in a single query instead of one
+        DB round-trip per candidate skill.
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def list_always_on_names(self) -> set[str]:
+        """Return the set of skill names marked 常驻(免压缩) (``always_on = True``).
+
+        Batch query used by the prompt builder to decide which skills bypass
+        skill compression — one query instead of a per-skill round-trip.
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    async def list_favorite_skill_names(self, user_id: str) -> set[str]:
+        """Return the skill names in ``user_id``'s "我的技能" collection.
+
+        These are skills the user does not own but that landed in their list —
+        currently only via their 岗位 (position) sync (``origin='position'``).
+        Owned skills are NOT included here (the caller detects those via
+        ``owner_user_id``).
+        """
+        raise NotImplementedError

offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/memory.py

@@ -0,0 +1,173 @@
+"""In-memory ``SkillStore`` implementation for tests."""
+
+from __future__ import annotations
+
+import copy
+from datetime import UTC, datetime
+from typing import Any
+
+from deerflow.persistence.skills.base import SkillStore
+
+
+def _now() -> str:
+    return datetime.now(UTC).isoformat()
+
+
+def _copy(record: dict[str, Any] | None) -> dict[str, Any] | None:
+    return copy.deepcopy(record) if record is not None else None
+
+
+def _with_derived(record: dict[str, Any]) -> dict[str, Any]:
+    out = dict(record)
+    out["builtin"] = out.get("owner_user_id") is None
+    out.setdefault("always_on", False)
+    return out
+
+
+class MemorySkillStore(SkillStore):
+    def __init__(self) -> None:
+        self._skills: dict[str, dict[str, Any]] = {}
+        # user_id -> {skill_name}
+        self._favorites: dict[str, set[str]] = {}
+
+    @staticmethod
+    def _is_visible(record: dict[str, Any], user_id: str) -> bool:
+        owner = record.get("owner_user_id")
+        if owner is None:
+            return True
+        if owner == user_id:
+            return True
+        return bool(record.get("published"))
+
+    async def list_visible(self, user_id: str) -> list[dict[str, Any]]:
+        return [
+            _with_derived(_copy(record))  # type: ignore[arg-type]
+            for record in self._skills.values()
+            if self._is_visible(record, user_id)
+        ]
+
+    async def get_visible(self, name: str, user_id: str) -> dict[str, Any] | None:
+        record = self._skills.get(name)
+        if record is None or not self._is_visible(record, user_id):
+            return None
+        return _with_derived(_copy(record))  # type: ignore[arg-type]
+
+    async def get_any(self, name: str) -> dict[str, Any] | None:
+        record = self._skills.get(name)
+        if record is None:
+            return None
+        return _with_derived(_copy(record))  # type: ignore[arg-type]
+
+    async def get_owned(self, name: str, user_id: str) -> dict[str, Any] | None:
+        record = self._skills.get(name)
+        if record is None or record.get("owner_user_id") != user_id:
+            return None
+        return _with_derived(_copy(record))  # type: ignore[arg-type]
+
+    async def create(self, data: dict[str, Any]) -> dict[str, Any]:
+        name = data["name"]
+        if name in self._skills:
+            raise ValueError(f"Skill '{name}' already exists")
+        record = {
+            "name": name,
+            "owner_user_id": data.get("owner_user_id"),
+            "published": bool(data.get("published", False)),
+            "detail": data.get("detail"),
+            "name_zh": data.get("name_zh"),
+            "always_on": bool(data.get("always_on", False)),
+            "square_id": str(data.get("square_id") or ""),
+            "created_at": _now(),
+            "updated_at": _now(),
+        }
+        self._skills[name] = record
+        return _with_derived(_copy(record))  # type: ignore[arg-type]
+
+    async def ensure_legacy(self, data: dict[str, Any]) -> dict[str, Any]:
+        name = data["name"]
+        record = self._skills.get(name)
+        if record is None:
+            record = {
+                "name": name,
+                "owner_user_id": data.get("owner_user_id"),
+                "published": bool(data.get("published", True)),
+                "detail": data.get("detail"),
+                "name_zh": data.get("name_zh"),
+                "created_at": _now(),
+                "updated_at": _now(),
+            }
+            self._skills[name] = record
+        elif record.get("owner_user_id") is None and data.get("owner_user_id") is not None:
+            record["owner_user_id"] = data["owner_user_id"]
+            record["updated_at"] = _now()
+        return _with_derived(_copy(record))  # type: ignore[arg-type]
+
+    async def update(self, name: str, user_id: str, data: dict[str, Any]) -> dict[str, Any] | None:
+        record = self._skills.get(name)
+        if record is None or record.get("owner_user_id") != user_id:
+            return None
+        if "published" in data:
+            record["published"] = bool(data["published"])
+        if "detail" in data:
+            record["detail"] = data["detail"]
+        if "name_zh" in data:
+            record["name_zh"] = data["name_zh"]
+        if "always_on" in data:
+            record["always_on"] = bool(data["always_on"])
+        if "square_id" in data:
+            record["square_id"] = str(data["square_id"] or "")
+        record["updated_at"] = _now()
+        return _with_derived(_copy(record))  # type: ignore[arg-type]
+
+    async def update_admin(self, name: str, data: dict[str, Any]) -> dict[str, Any] | None:
+        record = self._skills.get(name)
+        if record is None:
+            return None
+        if "published" in data:
+            record["published"] = bool(data["published"])
+        if "owner_user_id" in data:
+            record["owner_user_id"] = data["owner_user_id"]
+        if "detail" in data:
+            record["detail"] = data["detail"]
+        if "name_zh" in data:
+            record["name_zh"] = data["name_zh"]
+        if "always_on" in data:
+            record["always_on"] = bool(data["always_on"])
+        if "square_id" in data:
+            record["square_id"] = str(data["square_id"] or "")
+        record["updated_at"] = _now()
+        return _with_derived(_copy(record))  # type: ignore[arg-type]
+
+    async def delete(self, name: str, user_id: str) -> bool:
+        record = self._skills.get(name)
+        if record is None or record.get("owner_user_id") != user_id:
+            return False
+        del self._skills[name]
+        return True
+
+    async def delete_admin(self, name: str) -> bool:
+        if name not in self._skills:
+            return False
+        del self._skills[name]
+        return True
+
+    async def count_owned(self, user_id: str) -> int:
+        return sum(1 for s in self._skills.values() if s.get("owner_user_id") == user_id)
+
+    async def list_all(self) -> list[dict[str, Any]]:
+        return [
+            _with_derived(_copy(record))  # type: ignore[arg-type]
+            for record in self._skills.values()
+        ]
+
+    async def list_published_names(self) -> set[str]:
+        return {name for name, record in self._skills.items() if record.get("published")}
+
+    async def list_always_on_names(self) -> set[str]:
+        return {name for name, record in self._skills.items() if record.get("always_on")}
+
+    async def list_favorite_skill_names(self, user_id: str) -> set[str]:
+        return set(self._favorites.get(user_id, set()))
+
+    # test helper — not part of the abstract API
+    def add_favorite(self, user_id: str, skill_name: str) -> None:
+        self._favorites.setdefault(user_id, set()).add(skill_name)

offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/model.py

@@ -0,0 +1,85 @@
+"""ORM model for custom skill ownership + publish state.
+
+Note: skill content lives on disk under ``skills/custom/{name}/``; this table
+only stores the metadata (owner, published flag, timestamps). The skill
+``name`` doubles as the primary key because the filesystem path is already
+keyed by name, and we require global name uniqueness across all owners.
+"""
+
+from __future__ import annotations
+
+from datetime import UTC, datetime
+
+from sqlalchemy import Boolean, DateTime, String, Text, UniqueConstraint
+from sqlalchemy.orm import Mapped, mapped_column
+
+from deerflow.persistence.base import Base
+from deerflow.persistence.types import BeijingDateTime
+
+
+class SkillRow(Base):
+    __tablename__ = "skills"
+
+    # Global unique name — must match the on-disk folder name under skills/custom/.
+    # Hyphen-case or snake_case allowed; see deerflow.skills.validation.
+    name: Mapped[str] = mapped_column(String(64), primary_key=True)
+    owner_user_id: Mapped[str | None] = mapped_column(String(64), nullable=True, index=True)
+    published: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False, index=True)
+    # Human-authored long-form detail/introduction, editable by the skill owner
+    # or an admin. Distinct from the skill ``description`` parsed from SKILL.md;
+    # kept in the DB so admins can annotate built-in skills without rewriting
+    # the shared on-disk SKILL.md file.
+    detail: Mapped[str | None] = mapped_column(Text, nullable=True, default=None)
+    # User-configured or LLM-generated Chinese display name. Kept in the DB (not
+    # the on-disk SKILL.md) so it can be set for built-in skills too, and so the
+    # shared source file is never rewritten.
+    name_zh: Mapped[str | None] = mapped_column(String(128), nullable=True, default=None)
+    # 常驻(免压缩):开启技能压缩后,标记此项的技能仍以全文形式注入系统提示词,而非
+    # 降级为索引 + search_skills 检索。存 DB(而非 extensions_config.json),内置技能
+    # 也能标记,且避免在某些部署里回写源码目录下的配置文件导致报错。
+    always_on: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False, server_default="0", index=True)
+    # Publish square (发布广场) this skill is filed under once published. Pure
+    # organizational category, no access control. Empty string = unassigned →
+    # the listing layer resolves it to the default square. One square per skill.
+    square_id: Mapped[str] = mapped_column(String(64), nullable=False, default="", server_default="", index=True)
+    created_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, default=lambda: datetime.now(UTC))
+    updated_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, default=lambda: datetime.now(UTC), onupdate=lambda: datetime.now(UTC))
+
+
+class SkillFavoriteRow(Base):
+    """A user's "我的技能" membership for a skill they don't own.
+
+    Mirrors :class:`deerflow.persistence.agents.model.AgentFavoriteRow`: lets a
+    skill land in a user's "我的" list without the user owning it. Skills are
+    keyed by ``name`` (the on-disk folder / DB primary key), so this row
+    references ``skill_name`` rather than an opaque id. One row per
+    (user, skill); the unique constraint makes the grant idempotent.
+    """
+
+    __tablename__ = "skill_favorites"
+    __table_args__ = (UniqueConstraint("user_id", "skill_name", name="uq_skill_favorites_user_skill"),)
+
+    id: Mapped[str] = mapped_column(String(64), primary_key=True)
+    user_id: Mapped[str] = mapped_column(String(64), nullable=False, index=True)
+    skill_name: Mapped[str] = mapped_column(String(64), nullable=False, index=True)
+    # ``"user"`` = the user added it themselves (reserved for a future favorite
+    # button); ``"position"`` = auto-granted by the user's 岗位 (position) via
+    # the subscription sync engine. Sync only ever adds/removes ``"position"``
+    # rows, so a user's own picks are never clobbered.
+    origin: Mapped[str] = mapped_column(String(16), nullable=False, default="user", server_default="user")
+    # The position that granted this favorite (only set when ``origin='position'``).
+    position_id: Mapped[str | None] = mapped_column(String(64), nullable=True, index=True)
+    created_at: Mapped[datetime] = mapped_column(BeijingDateTime(), nullable=False, default=lambda: datetime.now(UTC))
+
+
+class SkillDisplayAdapterRow(Base):
+    __tablename__ = "skill_display_adapters"
+
+    skill_name: Mapped[str] = mapped_column(String(64), primary_key=True)
+    # Store JSON as text for maximum MySQL compatibility. Some deployments still
+    # run MySQL variants where native JSON columns or unquoted names such as
+    # ``display`` are not accepted consistently.
+    display_config: Mapped[str | None] = mapped_column(Text, nullable=True, default=None)
+    sample_payload: Mapped[str | None] = mapped_column(Text, nullable=True, default=None)
+    created_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, default=lambda: datetime.now(UTC))
+    updated_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, default=lambda: datetime.now(UTC), onupdate=lambda: datetime.now(UTC))

offline-backend-20260512/backend/packages/harness/deerflow/persistence/skills/sql.py

@@ -0,0 +1,190 @@
+"""SQLAlchemy-backed custom skill repository."""
+
+from __future__ import annotations
+
+from datetime import UTC, datetime
+from typing import Any
+
+from sqlalchemy import func, or_, select
+from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
+
+from deerflow.persistence.skills.base import SkillStore
+from deerflow.persistence.skills.model import SkillFavoriteRow, SkillRow
+
+
+def _row_to_dict(row: SkillRow) -> dict[str, Any]:
+    data = row.to_dict()
+    data["builtin"] = data.get("owner_user_id") is None
+    for key in ("created_at", "updated_at"):
+        value = data.get(key)
+        if isinstance(value, datetime):
+            data[key] = value.isoformat()
+    return data
+
+
+class SkillRepository(SkillStore):
+    def __init__(self, session_factory: async_sessionmaker[AsyncSession]) -> None:
+        self._sf = session_factory
+
+    @staticmethod
+    def _visible_filter(user_id: str):
+        return or_(
+            SkillRow.owner_user_id.is_(None),
+            SkillRow.owner_user_id == user_id,
+            SkillRow.published.is_(True),
+        )
+
+    async def list_visible(self, user_id: str) -> list[dict[str, Any]]:
+        stmt = select(SkillRow).where(self._visible_filter(user_id)).order_by(SkillRow.created_at.asc())
+        async with self._sf() as session:
+            result = await session.execute(stmt)
+            return [_row_to_dict(row) for row in result.scalars()]
+
+    async def get_visible(self, name: str, user_id: str) -> dict[str, Any] | None:
+        stmt = select(SkillRow).where(SkillRow.name == name, self._visible_filter(user_id))
+        async with self._sf() as session:
+            row = (await session.execute(stmt)).scalar_one_or_none()
+            return _row_to_dict(row) if row is not None else None
+
+    async def get_any(self, name: str) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(SkillRow, name)
+            return _row_to_dict(row) if row is not None else None
+
+    async def get_owned(self, name: str, user_id: str) -> dict[str, Any] | None:
+        stmt = select(SkillRow).where(SkillRow.name == name, SkillRow.owner_user_id == user_id)
+        async with self._sf() as session:
+            row = (await session.execute(stmt)).scalar_one_or_none()
+            return _row_to_dict(row) if row is not None else None
+
+    async def create(self, data: dict[str, Any]) -> dict[str, Any]:
+        now = datetime.now(UTC)
+        row = SkillRow(
+            name=data["name"],
+            owner_user_id=data.get("owner_user_id"),
+            published=bool(data.get("published", False)),
+            detail=data.get("detail"),
+            name_zh=data.get("name_zh"),
+            always_on=bool(data.get("always_on", False)),
+            square_id=str(data.get("square_id") or ""),
+            created_at=now,
+            updated_at=now,
+        )
+        async with self._sf() as session:
+            session.add(row)
+            await session.commit()
+            await session.refresh(row)
+            return _row_to_dict(row)
+
+    async def ensure_legacy(self, data: dict[str, Any]) -> dict[str, Any]:
+        async with self._sf() as session:
+            row = await session.get(SkillRow, data["name"])
+            if row is None:
+                now = datetime.now(UTC)
+                row = SkillRow(
+                    name=data["name"],
+                    owner_user_id=data.get("owner_user_id"),
+                    published=bool(data.get("published", True)),
+                    detail=data.get("detail"),
+                    name_zh=data.get("name_zh"),
+                    created_at=now,
+                    updated_at=now,
+                )
+                session.add(row)
+            elif row.owner_user_id is None and data.get("owner_user_id") is not None:
+                # Only patch ownership if it was previously null — never overwrite
+                # an existing owner record on a re-migration.
+                row.owner_user_id = data["owner_user_id"]
+                row.updated_at = datetime.now(UTC)
+            await session.commit()
+            await session.refresh(row)
+            return _row_to_dict(row)
+
+    async def update(self, name: str, user_id: str, data: dict[str, Any]) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(SkillRow, name)
+            if row is None or row.owner_user_id != user_id:
+                return None
+            if "published" in data:
+                row.published = bool(data["published"])
+            if "detail" in data:
+                row.detail = data["detail"]
+            if "name_zh" in data:
+                row.name_zh = data["name_zh"]
+            if "always_on" in data:
+                row.always_on = bool(data["always_on"])
+            if "square_id" in data:
+                row.square_id = str(data["square_id"] or "")
+            row.updated_at = datetime.now(UTC)
+            await session.commit()
+            await session.refresh(row)
+            return _row_to_dict(row)
+
+    async def update_admin(self, name: str, data: dict[str, Any]) -> dict[str, Any] | None:
+        async with self._sf() as session:
+            row = await session.get(SkillRow, name)
+            if row is None:
+                return None
+            if "published" in data:
+                row.published = bool(data["published"])
+            if "owner_user_id" in data:
+                row.owner_user_id = data["owner_user_id"]
+            if "detail" in data:
+                row.detail = data["detail"]
+            if "name_zh" in data:
+                row.name_zh = data["name_zh"]
+            if "always_on" in data:
+                row.always_on = bool(data["always_on"])
+            if "square_id" in data:
+                row.square_id = str(data["square_id"] or "")
+            row.updated_at = datetime.now(UTC)
+            await session.commit()
+            await session.refresh(row)
+            return _row_to_dict(row)
+
+    async def delete(self, name: str, user_id: str) -> bool:
+        async with self._sf() as session:
+            row = await session.get(SkillRow, name)
+            if row is None or row.owner_user_id != user_id:
+                return False
+            await session.delete(row)
+            await session.commit()
+            return True
+
+    async def delete_admin(self, name: str) -> bool:
+        async with self._sf() as session:
+            row = await session.get(SkillRow, name)
+            if row is None:
+                return False
+            await session.delete(row)
+            await session.commit()
+            return True
+
+    async def count_owned(self, user_id: str) -> int:
+        stmt = select(func.count()).select_from(SkillRow).where(SkillRow.owner_user_id == user_id)
+        async with self._sf() as session:
+            return await session.scalar(stmt) or 0
+
+    async def list_all(self) -> list[dict[str, Any]]:
+        stmt = select(SkillRow).order_by(SkillRow.created_at.asc())
+        async with self._sf() as session:
+            result = await session.execute(stmt)
+            return [_row_to_dict(row) for row in result.scalars()]
+
+    async def list_published_names(self) -> set[str]:
+        stmt = select(SkillRow.name).where(SkillRow.published.is_(True))
+        async with self._sf() as session:
+            result = await session.execute(stmt)
+            return {row[0] for row in result.all()}
+
+    async def list_always_on_names(self) -> set[str]:
+        stmt = select(SkillRow.name).where(SkillRow.always_on.is_(True))
+        async with self._sf() as session:
+            result = await session.execute(stmt)
+            return {row[0] for row in result.all()}
+
+    async def list_favorite_skill_names(self, user_id: str) -> set[str]:
+        stmt = select(SkillFavoriteRow.skill_name).where(SkillFavoriteRow.user_id == user_id)
+        async with self._sf() as session:
+            result = await session.execute(stmt)
+            return {row[0] for row in result.all()}

offline-backend-20260512/backend/packages/harness/deerflow/skills/__init__.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+from .installer import SkillAlreadyExistsError, SkillSecurityScanError
+from .storage import LocalSkillStorage, SkillStorage, get_or_new_skill_storage
+from .types import Skill
+from .validation import ALLOWED_FRONTMATTER_PROPERTIES, _validate_skill_frontmatter
+
+__all__ = [
+    "Skill",
+    "ALLOWED_FRONTMATTER_PROPERTIES",
+    "_validate_skill_frontmatter",
+    "SkillAlreadyExistsError",
+    "SkillSecurityScanError",
+    "SkillStorage",
+    "LocalSkillStorage",
+    "get_or_new_skill_storage",
+]

offline-backend-20260512/backend/packages/harness/deerflow/skills/address_scan.py

@@ -0,0 +1,255 @@
+"""Scan skill text for URL / IP addresses and apply safe batch replacements.
+
+Pure-logic layer for the 技能地址(URL/IP)批量替换 feature. This module knows
+**only about text** — it never walks the filesystem. The HTTP router
+(``app/gateway/routers/skill_addresses.py``) walks ``skills/{public,custom}``,
+reads each ``*.md`` / ``*.py`` file, and feeds the text here. Keeping the
+extraction / replacement logic here (no ``app.*`` import) makes it unit-testable
+in isolation and respects the harness/app boundary.
+
+Two address kinds are supported (the feature scope):
+
+* ``"url"`` — full ``http(s)://…`` URLs.
+* ``"ip"``  — dotted-quad IPv4, optionally ``:port``. Each octet is range
+  checked (0–255) so version-number noise like ``1.2.3.400`` is not matched.
+
+The single most important invariant is **substring-safe replacement**: we never
+``str.replace`` raw file text (that would rewrite ``1.2.3.4`` *inside*
+``11.2.3.40``). Instead we re-detect the exact token spans and only rewrite a
+span whose text is *exactly* one of the requested ``from`` values.
+"""
+
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+
+__all__ = [
+    "VALID_KINDS",
+    "AddressSpan",
+    "Occurrence",
+    "normalize_kinds",
+    "find_spans",
+    "scan_occurrences",
+    "count_matches",
+    "apply_replacements",
+    "is_valid_replacement",
+    "infer_kind",
+]
+
+# Supported address kinds (the configured feature scope: URL + IP).
+VALID_KINDS: tuple[str, ...] = ("url", "ip")
+
+# A URL body stops at whitespace, quotes, brackets, angle brackets, backticks
+# and backslashes — everything that commonly *delimits* a URL in md / py. The
+# trailing punctuation that often clings to the end (a sentence period, a
+# closing paren already excluded, a Markdown comma) is stripped afterwards.
+_URL_RE = re.compile(r"https?://[^\s'\"`<>)\]}\\]+", re.IGNORECASE)
+
+# Dotted-quad IPv4 (optionally :port). Guarded so it is not glued to a longer
+# number run (``(?<![\d.])`` / ``(?![\d.])``); octets are range-checked below.
+_IP_RE = re.compile(r"(?<![\d.])(\d{1,3}(?:\.\d{1,3}){3})(?::(\d{1,5}))?(?![\d.])")
+
+# Punctuation (ASCII + CJK) that should not be considered part of a URL token
+# when it appears at the very end of the match.
+_URL_TRAILING = set(".,;:!?'\")]}>`" + "）。，；：！？】》、…")
+
+
+@dataclass(frozen=True)
+class AddressSpan:
+    """An address token located at an exact ``[start, end)`` offset in text."""
+
+    kind: str
+    value: str
+    start: int
+    end: int
+
+
+@dataclass(frozen=True)
+class Occurrence:
+    """A located address enriched with 1-based line/column + the source line."""
+
+    kind: str
+    value: str
+    line: int
+    col: int
+    snippet: str
+
+
+def normalize_kinds(kinds: object) -> tuple[str, ...]:
+    """Coerce arbitrary input into an ordered, de-duped subset of ``VALID_KINDS``.
+
+    ``None`` / empty means "all kinds". Unknown kinds are dropped.
+    """
+
+    if kinds is None:
+        return VALID_KINDS
+    if isinstance(kinds, str):
+        raw = [p.strip() for p in kinds.split(",")]
+    else:
+        try:
+            raw = [str(p).strip() for p in kinds]  # type: ignore[assignment]
+        except TypeError:
+            return VALID_KINDS
+    seen: list[str] = []
+    for k in raw:
+        kl = k.lower()
+        if kl in VALID_KINDS and kl not in seen:
+            seen.append(kl)
+    return tuple(seen) if seen else VALID_KINDS
+
+
+def _valid_ipv4(quad: str) -> bool:
+    parts = quad.split(".")
+    if len(parts) != 4:
+        return False
+    for p in parts:
+        if not p.isdigit() or len(p) > 3:
+            return False
+        if int(p) > 255:
+            return False
+    return True
+
+
+def _valid_port(port: str | None) -> bool:
+    if port is None:
+        return True
+    return port.isdigit() and 0 <= int(port) <= 65535
+
+
+def _url_spans(text: str) -> list[AddressSpan]:
+    spans: list[AddressSpan] = []
+    for m in _URL_RE.finditer(text):
+        start, end = m.start(), m.end()
+        value = m.group(0)
+        # Strip clinging trailing punctuation.
+        while value and value[-1] in _URL_TRAILING:
+            value = value[:-1]
+            end -= 1
+        if value:
+            spans.append(AddressSpan(kind="url", value=value, start=start, end=end))
+    return spans
+
+
+def _ip_spans(text: str, exclude: list[AddressSpan]) -> list[AddressSpan]:
+    spans: list[AddressSpan] = []
+    for m in _IP_RE.finditer(text):
+        quad = m.group(1)
+        port = m.group(2)
+        if not _valid_ipv4(quad) or not _valid_port(port):
+            continue
+        start, end = m.start(), m.end()
+        # Skip an IP that lives *inside* a URL (e.g. http://1.2.3.4:8080/x) —
+        # replacing just the IP there would corrupt the URL.
+        if any(s.start <= start < s.end for s in exclude):
+            continue
+        spans.append(AddressSpan(kind="ip", value=m.group(0), start=start, end=end))
+    return spans
+
+
+def find_spans(text: str, kinds: object = None) -> list[AddressSpan]:
+    """Return every address token span in *text*, ordered by position.
+
+    URL regions are always computed first so that IPs embedded inside a URL are
+    not separately surfaced, even when only ``"ip"`` is requested.
+    """
+
+    wanted = normalize_kinds(kinds)
+    url_spans = _url_spans(text)
+    result: list[AddressSpan] = []
+    if "url" in wanted:
+        result.extend(url_spans)
+    if "ip" in wanted:
+        result.extend(_ip_spans(text, exclude=url_spans))
+    result.sort(key=lambda s: s.start)
+    return result
+
+
+def scan_occurrences(text: str, kinds: object = None, *, snippet_max: int = 240) -> list[Occurrence]:
+    """Locate addresses with 1-based line/column and a trimmed source line."""
+
+    spans = find_spans(text, kinds)
+    if not spans:
+        return []
+    # Precompute line-start offsets for O(log n) line lookup.
+    line_starts = [0]
+    for i, ch in enumerate(text):
+        if ch == "\n":
+            line_starts.append(i + 1)
+
+    occ: list[Occurrence] = []
+    import bisect
+
+    for s in spans:
+        line_idx = bisect.bisect_right(line_starts, s.start) - 1
+        line_no = line_idx + 1
+        col = s.start - line_starts[line_idx] + 1
+        line_end = line_starts[line_idx + 1] - 1 if line_idx + 1 < len(line_starts) else len(text)
+        raw_line = text[line_starts[line_idx]:line_end].rstrip("\r")
+        snippet = raw_line.strip()
+        if len(snippet) > snippet_max:
+            snippet = snippet[: snippet_max - 1] + "…"
+        occ.append(Occurrence(kind=s.kind, value=s.value, line=line_no, col=col, snippet=snippet))
+    return occ
+
+
+def count_matches(text: str, mapping: dict[str, str], kinds: object = None) -> int:
+    """How many address spans in *text* would be rewritten by *mapping*."""
+
+    if not mapping:
+        return 0
+    return sum(1 for s in find_spans(text, kinds) if s.value in mapping)
+
+
+def apply_replacements(text: str, mapping: dict[str, str], kinds: object = None) -> tuple[str, int]:
+    """Substring-safe batch replace.
+
+    Only spans whose text is *exactly* a key of *mapping* are rewritten, applied
+    right-to-left so earlier offsets stay valid. Returns ``(new_text, count)``.
+    A no-op mapping (or no hits) returns the original text and ``0``.
+    """
+
+    if not mapping:
+        return text, 0
+    spans = [s for s in find_spans(text, kinds) if s.value in mapping]
+    if not spans:
+        return text, 0
+    out = text
+    for s in sorted(spans, key=lambda s: s.start, reverse=True):
+        out = out[: s.start] + mapping[s.value] + out[s.end:]
+    return out, len(spans)
+
+
+def infer_kind(value: str) -> str | None:
+    """Return the address kind of *value* if it is a single, whole address.
+
+    Used to classify a user-supplied ``from`` value (so the matching ``to`` can
+    be validated against the right rules). Returns ``None`` when *value* is not
+    exactly one URL or IP token.
+    """
+
+    if not value:
+        return None
+    spans = find_spans(value, VALID_KINDS)
+    if len(spans) == 1 and spans[0].value == value:
+        return spans[0].kind
+    return None
+
+
+def is_valid_replacement(kind: str, value: str) -> bool:
+    """Validate that a user-supplied replacement is well-formed for its kind.
+
+    Keeps an admin from turning a URL into garbage (e.g. dropping the scheme) or
+    an IP into a non-IP. Empty replacements are rejected.
+    """
+
+    if not value or value.strip() != value or any(c.isspace() for c in value):
+        return False
+    if kind == "url":
+        return bool(re.fullmatch(r"https?://[^\s'\"`<>]+", value, re.IGNORECASE))
+    if kind == "ip":
+        m = _IP_RE.fullmatch(value)
+        if not m:
+            return False
+        return _valid_ipv4(m.group(1)) and _valid_port(m.group(2))
+    return False

offline-backend-20260512/backend/packages/harness/deerflow/skills/cascade.py

@@ -0,0 +1,206 @@
+"""Cascade handling when a published skill is removed from circulation.
+
+Called from the Gateway skills router whenever a published skill is
+unpublished, deleted, or admin-taken-down. Walks every agent on disk that
+referenced the skill, strips the skill from its ``config.yaml`` ``skills``
+list, and emits an in-app notification to that agent's owner.
+
+The harness/app firewall means we cannot import the FastAPI app side; the
+``AgentStore`` instance must be passed in by the caller (the router gets
+it from ``app.gateway.deps.get_agent_store``).
+"""
+
+from __future__ import annotations
+
+import logging
+from pathlib import Path
+from typing import Any, Literal
+
+import yaml
+
+from deerflow.config.agents_config import load_agent_config
+from deerflow.config.paths import get_paths
+from deerflow.persistence.notifications import notify
+
+logger = logging.getLogger(__name__)
+
+CascadeOperation = Literal["unpublish", "delete", "admin_takedown"]
+
+
+def _list_agent_dirs() -> list[Path]:
+    agents_dir = get_paths().agents_dir
+    if not agents_dir.exists():
+        return []
+    return [entry for entry in sorted(agents_dir.iterdir()) if entry.is_dir()]
+
+
+def _remove_skill_from_yaml(config_file: Path, skill_name: str) -> bool:
+    """Rewrite ``config.yaml`` removing ``skill_name`` from its ``skills`` list.
+
+    Returns True if the file was modified, False otherwise.
+    """
+    if not config_file.exists():
+        return False
+    try:
+        with config_file.open("r", encoding="utf-8") as f:
+            data: dict[str, Any] = yaml.safe_load(f) or {}
+    except yaml.YAMLError:
+        logger.warning("Cascade: failed to parse %s, skipping", config_file)
+        return False
+
+    skills = data.get("skills")
+    if not isinstance(skills, list) or skill_name not in skills:
+        return False
+
+    data["skills"] = [s for s in skills if s != skill_name]
+    with config_file.open("w", encoding="utf-8") as f:
+        yaml.safe_dump(data, f, default_flow_style=False, allow_unicode=True)
+    return True
+
+
+_OP_LABELS: dict[CascadeOperation, str] = {
+    "unpublish": "取消发布",
+    "delete": "删除",
+    "admin_takedown": "管理员强制下架",
+}
+
+
+async def cascade_skill_unpublished_or_deleted(
+    skill_name: str,
+    *,
+    actor_user_id: str,
+    operation: CascadeOperation,
+    skill_owner_user_id: str | None,
+    agent_store: Any,
+) -> dict[str, Any]:
+    """Detach ``skill_name`` from every referencing agent and notify owners.
+
+    Self-references (agent.user_id == actor_user_id) are still detached but
+    the owner gets no in-app notification — the user just did the action,
+    they know.
+
+    On ``admin_takedown`` we additionally notify ``skill_owner_user_id``.
+
+    Returns counters for testability: ``{checked, modified, notified}``.
+    """
+    counters = {"checked": 0, "modified": 0, "notified": 0}
+    label = _OP_LABELS.get(operation, operation)
+
+    for agent_dir in _list_agent_dirs():
+        config_file = agent_dir / "config.yaml"
+        if not config_file.exists():
+            continue
+        counters["checked"] += 1
+
+        # Quick read to decide whether we need to mutate at all.
+        try:
+            agent_cfg = load_agent_config(agent_dir.name)
+        except Exception:
+            logger.debug("Cascade: failed to load %s, skipping", agent_dir.name, exc_info=True)
+            continue
+        if agent_cfg is None or not agent_cfg.skills or skill_name not in agent_cfg.skills:
+            continue
+
+        modified = _remove_skill_from_yaml(config_file, skill_name)
+        if not modified:
+            continue
+        counters["modified"] += 1
+        logger.info(
+            "Cascade: removed skill '%s' from agent '%s/config.yaml' (op=%s)",
+            skill_name,
+            agent_dir.name,
+            operation,
+        )
+
+        # Look up agent owner from DB to decide where to send the notification.
+        record = await agent_store.get_any(agent_dir.name)
+        agent_owner = record.get("user_id") if record else None
+        if agent_owner and agent_owner != actor_user_id:
+            agent_display = (record.get("name") if record else None) or agent_dir.name
+            await notify(
+                agent_owner,
+                type="agent_skill_removed",
+                title=f"专家 “{agent_display}” 的一个技能已失效",
+                body=(
+                    f"技能 “{skill_name}” 被 {label}，已从该专家的技能列表中自动移除。"
+                ),
+                payload={
+                    "skill_name": skill_name,
+                    "agent_id": agent_dir.name,
+                    "agent_name": agent_display,
+                    "operation": operation,
+                    "actor_user_id": actor_user_id,
+                },
+            )
+            counters["notified"] += 1
+
+    if operation == "admin_takedown" and skill_owner_user_id and skill_owner_user_id != actor_user_id:
+        await notify(
+            skill_owner_user_id,
+            type="skill_taken_down",
+            title=f"你的技能 “{skill_name}” 已被管理员下架",
+            body="管理员对该技能执行了强制下架。如有疑问请联系管理员。",
+            payload={
+                "skill_name": skill_name,
+                "operation": operation,
+                "actor_user_id": actor_user_id,
+            },
+        )
+        counters["notified"] += 1
+
+    return counters
+
+
+def sync_agent_skills_to_available(available_skill_names: set[str]) -> dict[str, Any]:
+    """Strip skill references that no longer exist on disk from every agent.
+
+    Walks every agent dir and rewrites its ``config.yaml`` ``skills`` list to
+    contain only names that appear in ``available_skill_names``. Used by the
+    admin "agent-skills sync" endpoint to repair historical drift — skills
+    deleted before the cascade existed, deleted via a path that skipped the
+    cascade, or removed from the filesystem manually.
+
+    Synchronous because it only touches local files and the admin endpoint
+    runs it in a worker thread.
+
+    Returns ``{checked, modified, removals: [{agent_id, removed_skills}]}``.
+    """
+    counters: dict[str, Any] = {"checked": 0, "modified": 0, "removals": []}
+
+    for agent_dir in _list_agent_dirs():
+        config_file = agent_dir / "config.yaml"
+        if not config_file.exists():
+            continue
+        counters["checked"] += 1
+
+        try:
+            with config_file.open("r", encoding="utf-8") as f:
+                data: dict[str, Any] = yaml.safe_load(f) or {}
+        except yaml.YAMLError:
+            logger.warning("Sync: failed to parse %s, skipping", config_file)
+            continue
+
+        skills = data.get("skills")
+        if not isinstance(skills, list) or not skills:
+            continue
+
+        kept: list[str] = []
+        removed: list[str] = []
+        for name in skills:
+            if isinstance(name, str) and name in available_skill_names:
+                kept.append(name)
+            else:
+                removed.append(name if isinstance(name, str) else str(name))
+
+        if not removed:
+            continue
+
+        data["skills"] = kept
+        with config_file.open("w", encoding="utf-8") as f:
+            yaml.safe_dump(data, f, default_flow_style=False, allow_unicode=True)
+
+        counters["modified"] += 1
+        counters["removals"].append({"agent_id": agent_dir.name, "removed_skills": removed})
+        logger.info("Sync: removed orphan skills %s from agent '%s'", removed, agent_dir.name)
+
+    return counters

offline-backend-20260512/backend/packages/harness/deerflow/skills/curator_log.py

@@ -0,0 +1,111 @@
+"""In-memory ring buffer holding the most recent curator run's execution log.
+
+Now sharded per-user — each ``user_id`` gets its own deque so users only see
+their own curator activity from the Web UI.
+
+The buffer is intentionally process-local and non-persistent. The Gateway
+exposes it over ``GET /api/curator/logs`` with the active user id resolved
+from the request context.
+
+All public functions swallow internal exceptions so a logging failure can
+never break the curator run itself.
+"""
+
+from __future__ import annotations
+
+import logging
+import threading
+from collections import deque
+from datetime import UTC, datetime
+from typing import Any
+
+logger = logging.getLogger(__name__)
+
+_MAX_RECORDS = 200
+# Reserved bucket key for "no user context" callers (legacy code paths or
+# tests). Real curator runs always supply a concrete user_id.
+_DEFAULT_BUCKET = "__default__"
+
+_lock = threading.Lock()
+_buffers: dict[str, deque[dict[str, Any]]] = {}
+
+
+def _bucket_key(user_id: str | None) -> str:
+    return user_id or _DEFAULT_BUCKET
+
+
+def _get_or_create(user_id: str | None) -> deque[dict[str, Any]]:
+    key = _bucket_key(user_id)
+    buffer = _buffers.get(key)
+    if buffer is None:
+        buffer = deque(maxlen=_MAX_RECORDS)
+        _buffers[key] = buffer
+    return buffer
+
+
+def _now_iso() -> str:
+    return datetime.now(UTC).isoformat()
+
+
+def start_new_run(user_id: str | None = None) -> None:
+    """Clear the user's buffer and emit a ``run_started`` record."""
+    try:
+        with _lock:
+            buffer = _get_or_create(user_id)
+            buffer.clear()
+            buffer.append({"ts": _now_iso(), "level": "info", "message": "run_started"})
+    except Exception:
+        logger.debug("curator_log.start_new_run failed", exc_info=True)
+
+
+def record(level: str, message: str, *, user_id: str | None = None, **fields: Any) -> None:
+    """Append a single log record to the user's buffer. Never raises."""
+    try:
+        entry: dict[str, Any] = {
+            "ts": _now_iso(),
+            "level": str(level or "info"),
+            "message": str(message),
+        }
+        if fields:
+            entry.update({k: v for k, v in fields.items() if k not in entry})
+        with _lock:
+            buffer = _get_or_create(user_id)
+            buffer.append(entry)
+    except Exception:
+        logger.debug("curator_log.record failed", exc_info=True)
+
+
+def get_recent(user_id: str | None = None) -> list[dict[str, Any]]:
+    """Return a snapshot of records for ``user_id``. Never raises."""
+    try:
+        with _lock:
+            buffer = _buffers.get(_bucket_key(user_id))
+            if buffer is None:
+                return []
+            return list(buffer)
+    except Exception:
+        logger.debug("curator_log.get_recent failed", exc_info=True)
+        return []
+
+
+def clear(user_id: str | None = None) -> None:
+    """Drop a user's records. With ``user_id=None`` drops the default bucket.
+
+    Use ``clear_all()`` for cross-user cleanup (tests).
+    """
+    try:
+        with _lock:
+            buffer = _buffers.get(_bucket_key(user_id))
+            if buffer is not None:
+                buffer.clear()
+    except Exception:
+        logger.debug("curator_log.clear failed", exc_info=True)
+
+
+def clear_all() -> None:
+    """Drop every user's records. Used by tests."""
+    try:
+        with _lock:
+            _buffers.clear()
+    except Exception:
+        logger.debug("curator_log.clear_all failed", exc_info=True)

offline-backend-20260512/backend/packages/harness/deerflow/skills/installer.py

@@ -0,0 +1,204 @@
+"""Shared skill archive installation logic.
+
+Pure business logic — no FastAPI/HTTP dependencies.
+Both Gateway and Client delegate to these functions.
+"""
+
+import asyncio
+import concurrent.futures
+import logging
+import posixpath
+import shutil
+import stat
+import zipfile
+from pathlib import Path, PurePosixPath, PureWindowsPath
+
+from deerflow.skills.security_scanner import scan_skill_content
+
+logger = logging.getLogger(__name__)
+
+_PROMPT_INPUT_DIRS = {"references", "templates"}
+_PROMPT_INPUT_SUFFIXES = frozenset({".json", ".markdown", ".md", ".rst", ".txt", ".yaml", ".yml"})
+
+
+class SkillAlreadyExistsError(ValueError):
+    """Raised when a skill with the same name is already installed."""
+
+
+class SkillSecurityScanError(ValueError):
+    """Raised when a skill archive fails security scanning."""
+
+
+def is_unsafe_zip_member(info: zipfile.ZipInfo) -> bool:
+    """Return True if the zip member path is absolute or attempts directory traversal."""
+    name = info.filename
+    if not name:
+        return False
+    normalized = name.replace("\\", "/")
+    if normalized.startswith("/"):
+        return True
+    path = PurePosixPath(normalized)
+    if path.is_absolute():
+        return True
+    if PureWindowsPath(name).is_absolute():
+        return True
+    if ".." in path.parts:
+        return True
+    return False
+
+
+def is_symlink_member(info: zipfile.ZipInfo) -> bool:
+    """Detect symlinks based on the external attributes stored in the ZipInfo."""
+    mode = info.external_attr >> 16
+    return stat.S_ISLNK(mode)
+
+
+def should_ignore_archive_entry(path: Path) -> bool:
+    """Return True for macOS metadata dirs and dotfiles."""
+    return path.name.startswith(".") or path.name == "__MACOSX"
+
+
+def resolve_skill_dir_from_archive(temp_path: Path) -> Path:
+    """Locate the skill root directory from extracted archive contents.
+
+    Filters out macOS metadata (__MACOSX) and dotfiles (.DS_Store).
+
+    Returns:
+        Path to the skill directory.
+
+    Raises:
+        ValueError: If the archive is empty after filtering.
+    """
+    items = [p for p in temp_path.iterdir() if not should_ignore_archive_entry(p)]
+    if not items:
+        raise ValueError("Skill archive is empty")
+    if len(items) == 1 and items[0].is_dir():
+        return items[0]
+    return temp_path
+
+
+def safe_extract_skill_archive(
+    zip_ref: zipfile.ZipFile,
+    dest_path: Path,
+    max_total_size: int = 512 * 1024 * 1024,
+) -> None:
+    """Safely extract a skill archive with security protections.
+
+    Protections:
+    - Reject absolute paths and directory traversal (..).
+    - Skip symlink entries instead of materialising them.
+    - Enforce a hard limit on total uncompressed size (zip bomb defence).
+
+    Raises:
+        ValueError: If unsafe members or size limit exceeded.
+    """
+    dest_root = dest_path.resolve()
+    total_written = 0
+
+    for info in zip_ref.infolist():
+        if is_unsafe_zip_member(info):
+            raise ValueError(f"Archive contains unsafe member path: {info.filename!r}")
+
+        if is_symlink_member(info):
+            logger.warning("Skipping symlink entry in skill archive: %s", info.filename)
+            continue
+
+        normalized_name = posixpath.normpath(info.filename.replace("\\", "/"))
+        member_path = dest_root.joinpath(*PurePosixPath(normalized_name).parts)
+        if not member_path.resolve().is_relative_to(dest_root):
+            raise ValueError(f"Zip entry escapes destination: {info.filename!r}")
+        member_path.parent.mkdir(parents=True, exist_ok=True)
+
+        if info.is_dir():
+            member_path.mkdir(parents=True, exist_ok=True)
+            continue
+
+        with zip_ref.open(info) as src, member_path.open("wb") as dst:
+            while chunk := src.read(65536):
+                total_written += len(chunk)
+                if total_written > max_total_size:
+                    raise ValueError("Skill archive is too large or appears highly compressed.")
+                dst.write(chunk)
+
+
+def _is_script_support_file(rel_path: Path) -> bool:
+    return bool(rel_path.parts) and rel_path.parts[0] == "scripts"
+
+
+def _should_scan_support_file(rel_path: Path) -> bool:
+    if _is_script_support_file(rel_path):
+        return True
+    return bool(rel_path.parts) and rel_path.parts[0] in _PROMPT_INPUT_DIRS and rel_path.suffix.lower() in _PROMPT_INPUT_SUFFIXES
+
+
+def _move_staged_skill_into_reserved_target(staging_target: Path, target: Path) -> None:
+    installed = False
+    reserved = False
+    try:
+        target.mkdir(mode=0o700)
+        reserved = True
+        for child in staging_target.iterdir():
+            shutil.move(str(child), target / child.name)
+        installed = True
+    except FileExistsError as e:
+        raise SkillAlreadyExistsError(f"Skill '{target.name}' already exists") from e
+    finally:
+        if reserved and not installed and target.exists():
+            shutil.rmtree(target)
+
+
+async def _scan_skill_file_or_raise(skill_dir: Path, path: Path, skill_name: str, *, executable: bool) -> None:
+    rel_path = path.relative_to(skill_dir).as_posix()
+    location = f"{skill_name}/{rel_path}"
+    try:
+        content = path.read_text(encoding="utf-8")
+    except UnicodeDecodeError as e:
+        raise SkillSecurityScanError(f"Security scan failed for skill '{skill_name}': {location} must be valid UTF-8") from e
+
+    try:
+        result = await scan_skill_content(content, executable=executable, location=location)
+    except Exception as e:
+        raise SkillSecurityScanError(f"Security scan failed for {location}: {e}") from e
+
+    decision = getattr(result, "decision", None)
+    reason = str(getattr(result, "reason", "") or "No reason provided.")
+    if decision == "block":
+        if rel_path == "SKILL.md":
+            raise SkillSecurityScanError(f"Security scan blocked skill '{skill_name}': {reason}")
+        raise SkillSecurityScanError(f"Security scan blocked {location}: {reason}")
+    if executable and decision != "allow":
+        raise SkillSecurityScanError(f"Security scan rejected executable {location}: {reason}")
+    if decision not in {"allow", "warn"}:
+        raise SkillSecurityScanError(f"Security scan failed for {location}: invalid scanner decision {decision!r}")
+
+
+async def _scan_skill_archive_contents_or_raise(skill_dir: Path, skill_name: str) -> None:
+    """Run the skill security scanner against all installable text and script files."""
+    skill_md = skill_dir / "SKILL.md"
+    await _scan_skill_file_or_raise(skill_dir, skill_md, skill_name, executable=False)
+
+    for path in sorted(skill_dir.rglob("*")):
+        if not path.is_file():
+            continue
+
+        rel_path = path.relative_to(skill_dir)
+        if rel_path == Path("SKILL.md"):
+            continue
+        if path.name == "SKILL.md":
+            raise SkillSecurityScanError(f"Security scan failed for skill '{skill_name}': nested SKILL.md is not allowed at {skill_name}/{rel_path.as_posix()}")
+        if not _should_scan_support_file(rel_path):
+            continue
+
+        await _scan_skill_file_or_raise(skill_dir, path, skill_name, executable=_is_script_support_file(rel_path))
+
+
+def _run_async_install(coro):
+    try:
+        loop = asyncio.get_running_loop()
+    except RuntimeError:
+        loop = None
+
+    if loop is not None and loop.is_running():
+        with concurrent.futures.ThreadPoolExecutor(max_workers=1) as executor:
+            return executor.submit(asyncio.run, coro).result()
+    return asyncio.run(coro)

offline-backend-20260512/backend/packages/harness/deerflow/skills/parser.py

@@ -0,0 +1,80 @@
+import logging
+import re
+from pathlib import Path
+
+import yaml
+
+from .types import SKILL_MD_FILE, Skill, SkillCategory
+
+logger = logging.getLogger(__name__)
+
+
+def parse_skill_file(skill_file: Path, category: SkillCategory, relative_path: Path | None = None) -> Skill | None:
+    """Parse a SKILL.md file and extract metadata.
+
+    Args:
+        skill_file: Path to the SKILL.md file.
+        category: Category of the skill.
+        relative_path: Relative path from the category root to the skill
+            directory.  Defaults to the skill directory name when omitted.
+
+    Returns:
+        Skill object if parsing succeeds, None otherwise.
+    """
+    if not skill_file.exists() or skill_file.name != SKILL_MD_FILE:
+        return None
+
+    try:
+        content = skill_file.read_text(encoding="utf-8")
+
+        # Extract YAML front-matter block between leading ``---`` fences.
+        front_matter_match = re.match(r"^---\s*\n(.*?)\n---\s*\n", content, re.DOTALL)
+        if not front_matter_match:
+            return None
+
+        front_matter_text = front_matter_match.group(1)
+
+        try:
+            metadata = yaml.safe_load(front_matter_text)
+        except yaml.YAMLError as exc:
+            logger.error("Invalid YAML front-matter in %s: %s", skill_file, exc)
+            return None
+
+        if not isinstance(metadata, dict):
+            logger.error("Front-matter in %s is not a YAML mapping", skill_file)
+            return None
+
+        # Extract required fields.  Both must be non-empty strings.
+        name = metadata.get("name")
+        description = metadata.get("description")
+
+        if not name or not isinstance(name, str):
+            return None
+        if not description or not isinstance(description, str):
+            return None
+
+        # Normalise: strip surrounding whitespace that YAML may preserve.
+        name = name.strip()
+        description = description.strip()
+
+        if not name or not description:
+            return None
+
+        license_text = metadata.get("license")
+        if license_text is not None:
+            license_text = str(license_text).strip() or None
+
+        return Skill(
+            name=name,
+            description=description,
+            license=license_text,
+            skill_dir=skill_file.parent,
+            skill_file=skill_file,
+            relative_path=relative_path or Path(skill_file.parent.name),
+            category=category,
+            enabled=True,  # Actual state comes from the extensions config file.
+        )
+
+    except Exception:
+        logger.exception("Unexpected error parsing skill file %s", skill_file)
+        return None