diff --git a/frontend-web/src/components/workspace/agents/agent-card.tsx b/frontend-web/src/components/workspace/agents/agent-card.tsx
index 0842da4e..7a2c63fc 100644
--- a/frontend-web/src/components/workspace/agents/agent-card.tsx
+++ b/frontend-web/src/components/workspace/agents/agent-card.tsx
@@ -40,6 +40,7 @@ import { SkillDetailDialog } from "@/components/workspace/skills/skill-detail-di
 import { TagChipList } from "@/components/workspace/tags/tag-chip";
 import { TagPickerDialog } from "@/components/workspace/tags/tag-picker-dialog";
 import {
+  useAgent,
   useDeleteAgent,
   useSetAgentFeatured,
   useUpdateAgent,
@@ -74,6 +75,13 @@ export function AgentCard({ agent }: AgentCardProps) {
   );
   const [detailSkillName, setDetailSkillName] = useState<string | null>(null);
 
+  // The /api/agents list response intentionally omits `soul` (perf optimization
+  // — SOUL.md is only read on demand). Fetch the detail row when the edit
+  // dialog opens so the SOUL textarea can be hydrated.
+  const { agent: agentDetail, isLoading: isAgentDetailLoading } = useAgent(
+    editOpen ? agent.id : null,
+  );
+
   const storedAuth = useMemo(() => getStoredAuth(), []);
   const currentUserId = storedAuth?.user_id ?? null;
   const isAdmin = storedAuth?.system_role === "admin";
@@ -93,7 +101,6 @@ export function AgentCard({ agent }: AgentCardProps) {
     }
     setEditName(agent.name);
     setEditDescription(agent.description ?? "");
-    setEditSoul(agent.soul ?? "");
     setEditPublished(agent.published);
     setSelectedSkills(agent.skills ?? []);
   }, [
@@ -101,10 +108,23 @@ export function AgentCard({ agent }: AgentCardProps) {
     agent.name,
     agent.published,
     agent.skills,
-    agent.soul,
     editOpen,
   ]);
 
+  useEffect(() => {
+    if (!editOpen) {
+      return;
+    }
+    // SOUL.md only ships on the detail endpoint; fall back to whatever the list
+    // happened to carry so older deployments (or cached objects that still
+    // include `soul`) keep working.
+    if (agentDetail) {
+      setEditSoul(agentDetail.soul ?? "");
+    } else {
+      setEditSoul(agent.soul ?? "");
+    }
+  }, [agent.soul, agentDetail, editOpen]);
+
   function handleChat() {
     router.push(`/page/workspace/agents/${agent.id}/chats/new`);
   }
@@ -459,7 +479,11 @@ export function AgentCard({ agent }: AgentCardProps) {
               <Textarea
                 value={editSoul ?? ""}
                 onChange={(e) => setEditSoul(e.target.value)}
-                placeholder="输入 SOUL.md 内容"
+                placeholder={
+                  isAgentDetailLoading && !agentDetail
+                    ? "正在加载 SOUL.md..."
+                    : "输入 SOUL.md 内容"
+                }
                 className="min-h-40 font-mono text-xs"
               />
             </div>
diff --git a/frontend-web/src/pages/StrategyRoutes.tsx b/frontend-web/src/pages/StrategyRoutes.tsx
index 1fe204e5..0a83d16a 100644
--- a/frontend-web/src/pages/StrategyRoutes.tsx
+++ b/frontend-web/src/pages/StrategyRoutes.tsx
@@ -27,141 +27,162 @@ import TagsPage from "./TagsPage";
 import LightAppNewNotePage from "./LightAppNewNotePage";
 import LightAppNoteListPage from "./LightAppNoteListPage";
 import LightAppNoteQAPage from "./LightAppNoteQAPage";
+import TaskXkFrom from "@/strategy-components/components/task-components/TaskXkFrom";
+import TaskManagement from "@/strategy-components/components/task-components/TaskManagement";
 
 export default function StrategyRoutes() {
   return (
     <RouteBaseProvider value="/page/strategy">
-    <SidebarProvider>
-    <div className="h-full w-full overflow-hidden">
-      <Routes>
-        <Route index element={<Navigate to="qa/general" replace />} />
+      <SidebarProvider>
+        <div className="h-full w-full overflow-hidden">
+          <Routes>
+            <Route index element={<Navigate to="qa/general" replace />} />
 
-        {/* 问答中心 */}
-        <Route path="qa/general" element={<GeneralQA />} />
+            {/* 问答中心 */}
+            <Route path="qa/general" element={<GeneralQA />} />
+            {/* 任务管理 */}
+            <Route path="task-management" element={<TaskManagement />}>
+              {/* XK任务 */}
+              <Route
+                path="xk-task-jc"
+                element={
+                  <TaskXkFrom
+                    gopath={'#/pageIndie/pageHome/pageSwrw'}
+                  />
+                }
+              />
+              <Route
+                path="xk-task-dispatch"
+                element={
+                  <TaskXkFrom
+                    gopath={'#/pageIndie/pageJczl'}
+                  />
+                }
+              />
+            </Route>
+            {/* 轻应用 */}
+            <Route path="light-app/new-note" element={<LightAppNewNotePage />} />
+            <Route path="light-app/notes" element={<LightAppNoteListPage />} />
+            <Route path="light-app/note-qa" element={<LightAppNoteQAPage />} />
 
-        {/* 轻应用 */}
-        <Route path="light-app/new-note" element={<LightAppNewNotePage />} />
-        <Route path="light-app/notes" element={<LightAppNoteListPage />} />
-        <Route path="light-app/note-qa" element={<LightAppNoteQAPage />} />
+            {/* 对话 */}
+            <Route path="chats" element={<ChatsPage />} />
+            <Route
+              path="chats/:thread_id"
+              element={
+                <ChatRuntime>
+                  <ChatPage />
+                </ChatRuntime>
+              }
+            />
 
-        {/* 对话 */}
-        <Route path="chats" element={<ChatsPage />} />
-        <Route
-          path="chats/:thread_id"
-          element={
-            <ChatRuntime>
-              <ChatPage />
-            </ChatRuntime>
-          }
-        />
+            {/* 智能体管理 */}
+            <Route path="agents" element={<AgentsPage />} />
+            <Route path="agents/new" element={<NewAgentPage />} />
+            <Route
+              path="agents/:agent_id/chats"
+              element={<Navigate to="/page/strategy/agents" replace />}
+            />
+            <Route
+              path="agents/:agent_id/chats/:thread_id"
+              element={
+                <ChatRuntime>
+                  <AgentChatPage />
+                </ChatRuntime>
+              }
+            />
 
-        {/* 智能体管理 */}
-        <Route path="agents" element={<AgentsPage />} />
-        <Route path="agents/new" element={<NewAgentPage />} />
-        <Route
-          path="agents/:agent_id/chats"
-          element={<Navigate to="/page/strategy/agents" replace />}
-        />
-        <Route
-          path="agents/:agent_id/chats/:thread_id"
-          element={
-            <ChatRuntime>
-              <AgentChatPage />
-            </ChatRuntime>
-          }
-        />
+            {/* 技能 / 记忆 */}
+            <Route path="skills" element={<SkillsPage />} />
+            <Route path="skills/evolution" element={<SkillEvolutionPage />} />
+            <Route path="memory" element={<MemoryPage />} />
 
-        {/* 技能 / 记忆 */}
-        <Route path="skills" element={<SkillsPage />} />
-        <Route path="skills/evolution" element={<SkillEvolutionPage />} />
-        <Route path="memory" element={<MemoryPage />} />
+            {/* 定时任务 */}
+            <Route
+              path="scheduled"
+              element={<Navigate to="/page/strategy/scheduled/tasks" replace />}
+            />
+            <Route path="scheduled/tasks" element={<ScheduledTasksPage />} />
+            <Route
+              path="scheduled/runs/:run_id"
+              element={<ScheduledTaskRunDetailPage />}
+            />
 
-        {/* 定时任务 */}
-        <Route
-          path="scheduled"
-          element={<Navigate to="/page/strategy/scheduled/tasks" replace />}
-        />
-        <Route path="scheduled/tasks" element={<ScheduledTasksPage />} />
-        <Route
-          path="scheduled/runs/:run_id"
-          element={<ScheduledTaskRunDetailPage />}
-        />
-
-        {/* 设置 / 管理 — 权限与 workspace 侧一致，非 admin 重定向到 chats/new */}
-        <Route
-          path="settings/recommended-questions"
-          element={
-            getStoredAuth()?.system_role === "admin" ? (
-              <RecommendedQuestionsPage />
-            ) : (
-              <Navigate to="/page/workspace/chats/new" replace />
-            )
-          }
-        />
-        <Route
-          path="settings/prompt-prefix"
-          element={
-            getStoredAuth()?.system_role === "admin" ? (
-              <PromptPrefixSettingsPage />
-            ) : (
-              <Navigate to="/page/workspace/chats/new" replace />
-            )
-          }
-        />
-        <Route
-          path="settings/llm-metrics"
-          element={
-            getStoredAuth()?.system_role === "admin" ? (
-              <LlmMetricsPage />
-            ) : (
-              <Navigate to="/page/workspace/chats/new" replace />
-            )
-          }
-        />
-        <Route
-          path="admin/users"
-          element={
-            getStoredAuth()?.system_role === "admin" ? (
-              <UsersPage />
-            ) : (
-              <Navigate to="/page/workspace/chats/new" replace />
-            )
-          }
-        />
-        <Route
-          path="admin/scheduled-tasks"
-          element={
-            getStoredAuth()?.system_role === "admin" ? (
-              <AdminScheduledTasksPage />
-            ) : (
-              <Navigate to="/page/workspace/chats/new" replace />
-            )
-          }
-        />
-        <Route
-          path="admin/user-activity"
-          element={
-            getStoredAuth()?.system_role === "admin" ? (
-              <AdminUserActivityPage />
-            ) : (
-              <Navigate to="/page/workspace/chats/new" replace />
-            )
-          }
-        />
-        <Route
-          path="admin/tags"
-          element={
-            getStoredAuth()?.system_role === "admin" ? (
-              <TagsPage />
-            ) : (
-              <Navigate to="/page/workspace/chats/new" replace />
-            )
-          }
-        />
-      </Routes>
-    </div>
-    </SidebarProvider>
-    <Toaster position="top-center" />
+            {/* 设置 / 管理 — 权限与 workspace 侧一致，非 admin 重定向到 chats/new */}
+            <Route
+              path="settings/recommended-questions"
+              element={
+                getStoredAuth()?.system_role === "admin" ? (
+                  <RecommendedQuestionsPage />
+                ) : (
+                  <Navigate to="/page/workspace/chats/new" replace />
+                )
+              }
+            />
+            <Route
+              path="settings/prompt-prefix"
+              element={
+                getStoredAuth()?.system_role === "admin" ? (
+                  <PromptPrefixSettingsPage />
+                ) : (
+                  <Navigate to="/page/workspace/chats/new" replace />
+                )
+              }
+            />
+            <Route
+              path="settings/llm-metrics"
+              element={
+                getStoredAuth()?.system_role === "admin" ? (
+                  <LlmMetricsPage />
+                ) : (
+                  <Navigate to="/page/workspace/chats/new" replace />
+                )
+              }
+            />
+            <Route
+              path="admin/users"
+              element={
+                getStoredAuth()?.system_role === "admin" ? (
+                  <UsersPage />
+                ) : (
+                  <Navigate to="/page/workspace/chats/new" replace />
+                )
+              }
+            />
+            <Route
+              path="admin/scheduled-tasks"
+              element={
+                getStoredAuth()?.system_role === "admin" ? (
+                  <AdminScheduledTasksPage />
+                ) : (
+                  <Navigate to="/page/workspace/chats/new" replace />
+                )
+              }
+            />
+            <Route
+              path="admin/user-activity"
+              element={
+                getStoredAuth()?.system_role === "admin" ? (
+                  <AdminUserActivityPage />
+                ) : (
+                  <Navigate to="/page/workspace/chats/new" replace />
+                )
+              }
+            />
+            <Route
+              path="admin/tags"
+              element={
+                getStoredAuth()?.system_role === "admin" ? (
+                  <TagsPage />
+                ) : (
+                  <Navigate to="/page/workspace/chats/new" replace />
+                )
+              }
+            />
+          </Routes>
+        </div>
+      </SidebarProvider>
+      <Toaster position="top-center" />
     </RouteBaseProvider>
   );
 }
diff --git a/frontend-web/src/roundtable-planning/components/MessageStepsCard.tsx b/frontend-web/src/roundtable-planning/components/MessageStepsCard.tsx
index faffcc91..03ad67a4 100644
--- a/frontend-web/src/roundtable-planning/components/MessageStepsCard.tsx
+++ b/frontend-web/src/roundtable-planning/components/MessageStepsCard.tsx
@@ -46,6 +46,8 @@ export interface MessageStepsCardProps {
   className?: string;
   /** 点击 write_file / present_files 等产物相关步骤时打开沙箱。 */
   onArtifactStepClick?: (step: BubbleStep) => void;
+  /** 把 agent_orchestration 派活的 agent id 解析成中文席位名(用于 chip 显示)。 */
+  resolveAgentName?: (idOrName: string) => string;
 }
 
 /** 工具调用阶段的 step 才进入折叠卡;非工具步骤(connecting/streaming/tool_result)归到「思考」段。 */
@@ -58,6 +60,7 @@ export function MessageStepsCard({
   streaming,
   className,
   onArtifactStepClick,
+  resolveAgentName,
 }: MessageStepsCardProps) {
   // 把步骤拆成「工具调用」与「思考过程」两类 —— 与主聊天 MessageGroup 的 `lastToolCallStep` /
   // `lastReasoningStep` 划分对齐。
@@ -131,12 +134,14 @@ export function MessageStepsCard({
                 step={s}
                 status="complete"
                 onArtifactStepClick={onArtifactStepClick}
+                resolveAgentName={resolveAgentName}
               />
             ))}
           <ToolStep
             step={lastToolStep}
             status={streaming && trailingReasoning.length === 0 ? "active" : "complete"}
             onArtifactStepClick={onArtifactStepClick}
+            resolveAgentName={resolveAgentName}
           />
         </ChainOfThoughtContent>
       )}
@@ -155,13 +160,15 @@ function ToolStep({
   step,
   status,
   onArtifactStepClick,
+  resolveAgentName,
 }: {
   step: BubbleStep;
   status: "complete" | "active";
   onArtifactStepClick?: (step: BubbleStep) => void;
+  resolveAgentName?: (idOrName: string) => string;
 }) {
   const display = getStepDisplay(step.key, step.label, step.args);
-  const argNode = renderStepArgs(step.key, step.args);
+  const argNode = renderStepArgs(step.key, step.args, resolveAgentName);
   const [, toolName] = step.key.split(":");
   const isArtifactStep =
     toolName === "write_file" ||
diff --git a/frontend-web/src/roundtable-planning/components/Step1Panel.tsx b/frontend-web/src/roundtable-planning/components/Step1Panel.tsx
index 3dd6f1de..6184e20d 100644
--- a/frontend-web/src/roundtable-planning/components/Step1Panel.tsx
+++ b/frontend-web/src/roundtable-planning/components/Step1Panel.tsx
@@ -353,7 +353,13 @@ function Step1Message({
         <div className="text-xs text-slate-700 leading-relaxed font-medium break-words space-y-2 roundtable-md">
           {introText && <MarkdownContent>{introText}</MarkdownContent>}
           {parts.summary && <IntentReadySummaryCard summary={parts.summary} />}
-          {parts.after && <MarkdownContent>{parts.after}</MarkdownContent>}
+          {/*
+            parts.after 是模型违反 SOUL §0「硬规则 3」在 [INTENT_READY] 块后
+            继续输出的内容(典型:"那么我先帮你列一下大纲……")。后端 intent.py
+            的终态帧已经把 content 截断到 marker 结尾,但流式期间 LangGraph
+            messages-tuple 还是会把那段废话逐 token 推过来——故意在 UI 层
+            把它丢弃,让用户看到的就是"复述 + 摘要卡",干净结束。
+          */}
           {isStreaming && <StreamingCursor />}
           {isPendingClarification && clarification && (
             <IntentClarificationCard
diff --git a/frontend-web/src/roundtable-planning/components/Step2Panel.tsx b/frontend-web/src/roundtable-planning/components/Step2Panel.tsx
index a0a947ec..76dc50b5 100644
--- a/frontend-web/src/roundtable-planning/components/Step2Panel.tsx
+++ b/frontend-web/src/roundtable-planning/components/Step2Panel.tsx
@@ -67,6 +67,11 @@ export interface Step2PanelProps {
   agentConfigs: Record<number, AgentConfig>;
   selectedAgentId: number | null;
   roles: RoleEntry[];
+  /** agent id(小写)→ 席位元数据,用于把派活 chip 的 id 显示成中文席位名。 */
+  agentNameToRole: Record<
+    string,
+    { speakerId: number; avatarType: string; displayName: string }
+  >;
   selectedModel: string;
   availableModels: ModelOption[];
   modelsLoading: boolean;
@@ -109,6 +114,7 @@ export function Step2Panel(props: Step2PanelProps) {
     agentConfigs,
     selectedAgentId,
     roles,
+    agentNameToRole,
     selectedModel,
     availableModels,
     modelsLoading,
@@ -134,6 +140,12 @@ export function Step2Panel(props: Step2PanelProps) {
   const { atBottom: step2AtBottom, scrollToBottom: scrollStep2ToBottom } =
     useAutoScroll(step2ScrollRef, [step2RoundtableDialogues]);
 
+  const resolveAgentName = useCallback(
+    (idOrName: string) =>
+      agentNameToRole[idOrName.toLowerCase()]?.displayName ?? idOrName,
+    [agentNameToRole],
+  );
+
   // 干预研审:记录指令并自动恢复会商,右上角按钮状态由 isRoundTableRunning 驱动
   const [pauseFlash, setPauseFlash] = useState(false);
 
@@ -245,6 +257,7 @@ export function Step2Panel(props: Step2PanelProps) {
             <Step2Message
               key={msg.id}
               msg={msg}
+              resolveAgentName={resolveAgentName}
               onArtifactStepClick={(step) => openArtifactFromStep(msg, step)}
               onFileCardClick={(path) => {
                 if (msg.agentThreadId) {
@@ -396,6 +409,8 @@ export function Step2Panel(props: Step2PanelProps) {
  */
 interface Step2MessageProps {
   msg: Step2Dialogue;
+  /** 把派活 chip 的 agent id 解析成中文席位名。 */
+  resolveAgentName?: (idOrName: string) => string;
   onArtifactStepClick?: (
     step: NonNullable<Step2Dialogue["steps"]>[number],
   ) => void;
@@ -405,6 +420,7 @@ interface Step2MessageProps {
 
 function Step2Message({
   msg,
+  resolveAgentName,
   onArtifactStepClick,
   onFileCardClick,
 }: Step2MessageProps) {
@@ -435,6 +451,7 @@ function Step2Message({
           steps={msg.steps}
           streaming={isStreaming}
           onArtifactStepClick={onArtifactStepClick}
+          resolveAgentName={resolveAgentName}
         />
       )}
       {reasoning && (
diff --git a/frontend-web/src/roundtable-planning/hooks/useStep2Orchestration.ts b/frontend-web/src/roundtable-planning/hooks/useStep2Orchestration.ts
index 2def91ff..7e4f7250 100644
--- a/frontend-web/src/roundtable-planning/hooks/useStep2Orchestration.ts
+++ b/frontend-web/src/roundtable-planning/hooks/useStep2Orchestration.ts
@@ -434,6 +434,51 @@ export function useStep2Orchestration(
     [],
   );
 
+  /**
+   * 反查某个产出文件是由哪个席位 thread 写入的。
+   *
+   * 总控(coordinator)常用 `present_files` 呈现「子智能体写的文件」,但文件物理上落在
+   * 子席位 thread 的 outputs 里,coordinator thread 并没有 —— 直接拿总控气泡的
+   * `agentThreadId`(=coordinator thread)走 HTTP 拉会 404(`Thread ... not found`)。
+   * 流式期间子席位 write_file 时,内容已缓存进 seatStubMessages(key=真实写入 thread)。
+   * 这里按 path 反查那条 stub,让关闭沙箱后再点文件卡也能复用虚拟预览(内存内容,
+   * 与流式期间体验一致,且不依赖后端 thread 是否仍可访问)。
+   */
+  const findSeatArtifactSource = useCallback(
+    (
+      rawPath: string,
+    ): { seatThreadId: string; messageId: string; toolCallId: string } | null => {
+      const target = rawPath.trim();
+      if (!target) return null;
+      const targetBase = target.split("/").pop() ?? target;
+      let basenameMatch:
+        | { seatThreadId: string; messageId: string; toolCallId: string }
+        | null = null;
+      for (const [seatThreadId, msgs] of Object.entries(
+        seatStubMessagesRef.current,
+      )) {
+        const msg = msgs?.[0];
+        if (!msg) continue;
+        for (const tc of msg.tool_calls) {
+          if (tc.name !== "write_file" && tc.name !== "str_replace") continue;
+          const p =
+            typeof (tc.args as { path?: unknown })?.path === "string"
+              ? ((tc.args as { path?: string }).path as string)
+              : "";
+          if (!p) continue;
+          if (p === target) {
+            return { seatThreadId, messageId: msg.id, toolCallId: tc.id };
+          }
+          if (!basenameMatch && (p.split("/").pop() ?? p) === targetBase) {
+            basenameMatch = { seatThreadId, messageId: msg.id, toolCallId: tc.id };
+          }
+        }
+      }
+      return basenameMatch;
+    },
+    [],
+  );
+
   /**
    * `openArtifactPreview` —— 走真实路径(present_files 等已落盘场景),走 HTTP API 加载。
    * 沙箱 write_file 实时预览请用 `openSeatVirtualArtifact`。
@@ -441,7 +486,27 @@ export function useStep2Orchestration(
   const openArtifactPreview = useCallback(
     (threadId: string, path: string, options?: { auto?: boolean }) => {
       const normalized = path.trim();
-      if (!normalized || !threadId) return;
+      if (!normalized) return;
+
+      // present_files 呈现的文件多半是子智能体 write_file 写入的:真实内容/thread 在 stub 里。
+      // 先反查 stub 走虚拟预览(内存内容)—— 关闭沙箱后再点文件卡也能显示,且不受
+      // 总控 coordinator thread 在后端 artifacts 路由下 not found 的影响。
+      const seatSource = findSeatArtifactSource(normalized);
+      if (seatSource) {
+        const virtualUrl = buildWriteFileVirtualUrl(
+          normalized,
+          seatSource.messageId,
+          seatSource.toolCallId,
+        );
+        openSeatVirtualArtifact(
+          seatSource.seatThreadId,
+          virtualUrl,
+          options?.auto ?? false,
+        );
+        return;
+      }
+
+      if (!threadId) return;
 
       setActiveArtifactThreadId(threadId);
       const prev = artifactPathsRef.current;
@@ -462,7 +527,14 @@ export function useStep2Orchestration(
         openPanel();
       }
     },
-    [selectArtifact, setArtifacts, setArtifactsOpen],
+    [
+      findSeatArtifactSource,
+      buildWriteFileVirtualUrl,
+      openSeatVirtualArtifact,
+      selectArtifact,
+      setArtifacts,
+      setArtifactsOpen,
+    ],
   );
 
   const openArtifactFromStep = useCallback(
diff --git a/frontend-web/src/roundtable-planning/lib/roundtable-constants.ts b/frontend-web/src/roundtable-planning/lib/roundtable-constants.ts
index 37e90e31..1fefa30e 100644
--- a/frontend-web/src/roundtable-planning/lib/roundtable-constants.ts
+++ b/frontend-web/src/roundtable-planning/lib/roundtable-constants.ts
@@ -140,6 +140,75 @@ export function newCoordinatorName(): string {
  */
 export const INTENT_READY_RE = /\[INTENT_READY\]\s*```json\s*(\{[\s\S]*?\})\s*```/;
 
+/**
+ * 把 JSON 文本里**字符串值内部**的字面控制字符(换行 / 回车 / 制表符)转义成合法的
+ * `\n` / `\r` / `\t`。模型偶尔会在 summary 的字符串值里直接敲换行,产出的 JSON 不合法,
+ * `JSON.parse` 会抛 `Bad control character in string literal`。这里只在字符串内替换,
+ * 不动结构性空白,避免破坏本来就合法的 JSON。
+ */
+function sanitizeJsonControlChars(raw: string): string {
+  let out = "";
+  let inString = false;
+  let escaped = false;
+  for (const ch of raw) {
+    if (escaped) {
+      out += ch;
+      escaped = false;
+      continue;
+    }
+    if (ch === "\\") {
+      out += ch;
+      escaped = true;
+      continue;
+    }
+    if (ch === '"') {
+      inString = !inString;
+      out += ch;
+      continue;
+    }
+    if (inString) {
+      if (ch === "\n") {
+        out += "\\n";
+        continue;
+      }
+      if (ch === "\r") {
+        out += "\\r";
+        continue;
+      }
+      if (ch === "\t") {
+        out += "\\t";
+        continue;
+      }
+    }
+    out += ch;
+  }
+  return out;
+}
+
+function parseIntentSummary(raw: string): IntentSummary | null {
+  const attempt = (s: string): IntentSummary | null => {
+    const parsed = JSON.parse(s) as unknown;
+    if (
+      parsed &&
+      typeof parsed === "object" &&
+      typeof (parsed as { objective?: unknown }).objective === "string"
+    ) {
+      return parsed as IntentSummary;
+    }
+    return null;
+  };
+  try {
+    return attempt(raw);
+  } catch {
+    // 字符串值里有裸换行 / 制表符 → 转义后重试,避免整块摘要丢失导致 Step 2 无法解锁。
+    try {
+      return attempt(sanitizeJsonControlChars(raw));
+    } catch {
+      return null;
+    }
+  }
+}
+
 /**
  * 把含 [INTENT_READY] 标记的气泡正文拆成前 / 摘要 / 后三段。
  * 摘要 JSON 格式异常时整体保留在 `before`,用户仍能看到原文。
@@ -152,15 +221,8 @@ export function splitIntentReadyBlock(text: string): {
   if (!text) return { before: text, summary: null, after: "" };
   const m = INTENT_READY_RE.exec(text);
   if (!m) return { before: text, summary: null, after: "" };
-  let summary: IntentSummary | null = null;
-  try {
-    const parsed = JSON.parse(m[1]);
-    if (parsed && typeof parsed === "object" && typeof parsed.objective === "string") {
-      summary = parsed as IntentSummary;
-    }
-  } catch {
-    return { before: text, summary: null, after: "" };
-  }
+  const summary = parseIntentSummary(m[1]);
+  if (!summary) return { before: text, summary: null, after: "" };
   return {
     before: text.slice(0, m.index).trimEnd(),
     summary,
diff --git a/frontend-web/src/roundtable-planning/lib/step-display.tsx b/frontend-web/src/roundtable-planning/lib/step-display.tsx
index 87e7dabf..6f197cd9 100644
--- a/frontend-web/src/roundtable-planning/lib/step-display.tsx
+++ b/frontend-web/src/roundtable-planning/lib/step-display.tsx
@@ -23,6 +23,8 @@ import {
 } from "lucide-react";
 
 import { ChainOfThoughtSearchResult } from "@/components/ai-elements/chain-of-thought";
+import { CodeBlock } from "@/components/ai-elements/code-block";
+import { MarkdownContent } from "@/roundtable-planning/components/MessageBubble";
 
 /**
  * 根据工具名 + args 渲染 ChainOfThoughtStep 的子节点(路径 chip / 命令代码块等)。
@@ -34,6 +36,7 @@ import { ChainOfThoughtSearchResult } from "@/components/ai-elements/chain-of-th
 export function renderStepArgs(
   key: string,
   args?: Record<string, unknown>,
+  resolveAgentName?: (idOrName: string) => string,
 ): React.ReactNode {
   if (!args) return null;
   const [, toolName] = key.split(":");
@@ -43,15 +46,18 @@ export function renderStepArgs(
       const agentName = typeof args.agent_name === "string" ? args.agent_name : "";
       const task = typeof args.task === "string" ? args.task : "";
       if (!agentName && !task) return null;
+      // 派活 chip 显示中文席位名而非 agent id(客户自建 agent 的 id 常是一串 hex)。
+      const displayName =
+        agentName && resolveAgentName ? resolveAgentName(agentName) : agentName;
       return (
         <>
-          {agentName && (
-            <ChainOfThoughtSearchResult>{agentName}</ChainOfThoughtSearchResult>
+          {displayName && (
+            <ChainOfThoughtSearchResult>{displayName}</ChainOfThoughtSearchResult>
           )}
           {task && (
-            <span className="block text-[11px] text-slate-600 line-clamp-3 mt-0.5">
-              {task}
-            </span>
+            <div className="mt-1 max-h-64 overflow-y-auto rounded-md border border-slate-200 bg-white px-2.5 py-1 text-[11px] text-slate-600 leading-relaxed">
+              <MarkdownContent>{task}</MarkdownContent>
+            </div>
           )}
         </>
       );
@@ -75,9 +81,14 @@ export function renderStepArgs(
     case "bash": {
       const command = typeof args.command === "string" ? args.command : "";
       return command ? (
-        <code className="block bg-slate-100 rounded px-2 py-1 text-[11px] font-mono text-slate-800 mt-0.5 line-clamp-2 break-all">
-          {command}
-        </code>
+        <div className="mt-1 max-h-60 overflow-auto rounded-md border border-slate-200 [&_pre]:px-2.5! [&_pre]:py-1.5!">
+          <CodeBlock
+            className="rounded-none border-none"
+            showLineNumbers={false}
+            language="bash"
+            code={command}
+          />
+        </div>
       ) : null;
     }
     case "write_todos": {
diff --git a/frontend-web/src/roundtable-planning/pages/RoundtablePlanningPage.tsx b/frontend-web/src/roundtable-planning/pages/RoundtablePlanningPage.tsx
index d775b5d8..5195efce 100644
--- a/frontend-web/src/roundtable-planning/pages/RoundtablePlanningPage.tsx
+++ b/frontend-web/src/roundtable-planning/pages/RoundtablePlanningPage.tsx
@@ -488,6 +488,7 @@ export default function RoundtablePlanningPage() {
               agentConfigs={step2.agentConfigs}
               selectedAgentId={step2.selectedAgentId}
               roles={roles}
+              agentNameToRole={step2.agentNameToRole}
               selectedModel={selectedModel}
               availableModels={availableModels}
               modelsLoading={modelsLoading}
diff --git a/frontend-web/src/strategy-components/components/task-components/TaskManagement.tsx b/frontend-web/src/strategy-components/components/task-components/TaskManagement.tsx
new file mode 100644
index 00000000..588e6b45
--- /dev/null
+++ b/frontend-web/src/strategy-components/components/task-components/TaskManagement.tsx
@@ -0,0 +1,166 @@
+import { Outlet } from 'react-router-dom';
+export interface SceneInfo {
+  applicableTask: string;
+  createTime: string;
+  deleted: number;
+  id: number;
+  relatedAgent: string;
+  sceneCode: string;
+  sceneIntro: string;
+  sceneName: string;
+  sceneType: string;
+  status: number;
+  updateTime: string;
+  version: string;
+}
+
+export interface TaskListItem {
+  createBy: number;
+  createTime: string;
+  deptId: null;
+  id: number;
+  isTop: number;
+  issuingDept: string;
+  keywords: null;
+  overview: string;
+  planTaskTop: number;
+  platform: null;
+  ptaskId: null;
+  recommendedBasis: null;
+  scene: string; // 或 SceneInfo，如果你已经解析
+  sourceType: number;
+  step: null;
+  taskContent: string;
+  taskDirection: string;
+  taskEndDate: string;
+  taskLevel: number;
+  taskName: string;
+  taskNo: string;
+  taskStartDate: string;
+  taskStatus: number;
+  taskSuperiorFlag: number;
+  taskTimeLimit: null;
+  taskType: number;
+  column: string;
+  updateBy: number;
+  updateTime: string;
+  endCount: number;
+  sumAll: number;
+  progress: number;
+  taskStatusData: any;
+}
+
+export interface ActionListItem {
+  actionDay: number;
+  actionEndDate: string;
+  actionManager: number;
+  actionName: string;
+  actionNo: any;
+  actionStartDate: string;
+  actionStatus: number;
+  actionTerm: number;
+  actionTitle: any;
+  actorName: any;
+  audienceGroup: any
+  audienceGroupId: any;
+  audienceRecReason: any
+  behavior: any
+  createBy: number
+  createTime: string;
+  deleteFlag: number;
+  deptId: any;
+  deptName: string;
+  drivingFactor: any;
+  expectedBehaviorId: any;
+  id: number
+  keywords: string
+  level: number
+  scene: string
+  sourceType: number
+  step: null
+  taskId: number
+  updateBy: number
+  updateTime: number
+  useActionId: null
+}
+
+export interface actionColumnItem {
+  icon: string;
+  columnName: string;
+  actionList: ActionListItem[]
+}
+
+export const taskStatusList = {
+  '0': {
+    label: '待签收',
+    color: '#0062d9',
+    icon: '待签收.svg'
+  },
+  '10': {
+    label: '待分析',
+    color: '#2f9acc',
+    icon: '待CEH.svg'
+  },
+  '20': {
+    label: '分析中',
+    color: '#d97b00',
+    icon: '进行中.svg'
+  },
+  '25': {
+    label: '分析完成已提交',
+    color: '#1eb07f',
+    icon: '完成.svg'
+  },
+  '30': {
+    label: '已分析待审核',
+    color: '#1eb07f',
+    icon: '完成.svg'
+  },
+  '40': {
+    label: '分析审核通过',
+    color: '#1eb07f',
+    icon: '完成.svg'
+  },
+  '50': {
+    label: '分析审核未通过',
+    color: '#cf1322',
+    icon: '未通过.svg'
+  },
+  '60': {
+    label: '行动已下发待签收',
+    color: '#1eb07f',
+    icon: '完成.svg'
+  },
+  '70': {
+    label: '行动已签收待策划',
+    color: '#1eb07f',
+    icon: '完成.svg'
+  },
+  '80': {
+    label: '行动策划已提交',
+    color: '#1eb07f',
+    icon: '完成.svg'
+  },
+  '100': {
+    label: '作战方案待审核',
+    color: '#1eb07f',
+    icon: '完成.svg'
+  },
+  '110': {
+    label: '作战方案审核通过',
+    color: '#1eb07f',
+    icon: '完成.svg'
+  },
+  '120': {
+    label: '作战方案未通过',
+    color: '#cf1322',
+    icon: '未通过.svg'
+  }
+};
+export default function TaskManagement() {
+  return (
+    <div className="h-full flex flex-col">
+      <Outlet />
+    </div>
+  );
+}
diff --git a/frontend-web/src/strategy-components/components/task-components/TaskXkFrom.tsx b/frontend-web/src/strategy-components/components/task-components/TaskXkFrom.tsx
new file mode 100644
index 00000000..832dffa5
--- /dev/null
+++ b/frontend-web/src/strategy-components/components/task-components/TaskXkFrom.tsx
@@ -0,0 +1,23 @@
+import { Outlet, useNavigate } from 'react-router-dom';
+import React, { useState, useEffect, useRef } from 'react';
+import { useSearchParams } from 'react-router-dom';
+const userInfo = localStorage.getItem('userInfo');
+const xkLinkUrl = import.meta.env.VITE_AIPI_XKLINKTO_URL;
+interface TaskPgFromProps {
+  gopath: string
+}
+const TaskXkFrom: React.FC<TaskPgFromProps> = ({gopath}) => {
+const [xkUrl, setXkUrl] = useState(xkLinkUrl);
+  useEffect(() => {
+    if(userInfo) {
+      const userData: any = JSON.parse(userInfo);
+      setXkUrl(`${xkLinkUrl}${gopath}?accessToken=${userData.accessToken}`);
+    }
+  }, [gopath]);
+  return (
+    <div className="h-full flex flex-col">
+      <iframe className="h-[calc(100vh-92px)]" src={xkUrl}></iframe>
+    </div>
+  );
+}
+export default TaskXkFrom
diff --git a/offline-backend-20260512/backend/app/gateway/routers/intent.py b/offline-backend-20260512/backend/app/gateway/routers/intent.py
index f53ed6bd..1b9a5c1f 100644
--- a/offline-backend-20260512/backend/app/gateway/routers/intent.py
+++ b/offline-backend-20260512/backend/app/gateway/routers/intent.py
@@ -133,6 +133,70 @@ def _normalize_options(raw: object) -> tuple[str, ...]:
     return tuple(sorted(o for o in out if o))
 
 
+def _count_past_clarifications(messages: list[dict[str, Any]]) -> int:
+    """统计当前 thread 内已经发出过的 ``ask_clarification`` 次数。
+
+    用来给 ``stream_intent`` 做"3 轮硬上限"兜底:
+      - 0-1 次:正常流程,模型自行决定继续问还是收敛
+      - 2 次(即将进入第 3 轮):prompt 追加系统提示,要求本轮必须 INTENT_READY
+      - ≥3 次:工具白名单移除 ``ask_clarification``,模型物理上无法再问
+
+    只数"已经成型"的 tool_call,不区分用户是否回答过——LangGraph 把 ask
+    工具的调用永久记录在 AIMessage.tool_calls 里,数它即可。
+    """
+    if not isinstance(messages, list):
+        return 0
+    count = 0
+    for msg in messages:
+        if not isinstance(msg, dict):
+            continue
+        if (msg.get("type") or "").lower() not in ("ai", "aimessage", "aimessagechunk"):
+            continue
+        for tc in msg.get("tool_calls") or []:
+            if (tc.get("name") or "").lower() == "ask_clarification":
+                count += 1
+    return count
+
+
+async def _fetch_thread_messages(
+    client: httpx.AsyncClient,
+    base: str,
+    headers: dict[str, str],
+    thread_id: str,
+) -> list[dict[str, Any]]:
+    """读取当前 thread checkpoint 的 messages 列表(best-effort)。
+
+    用 ASGITransport loopback 调 ``GET /api/threads/{tid}/state``,失败时返回
+    空列表——计数兜底是个"刹车",失败时退化成原有的纯 SOUL 自律行为,不阻断主流程。
+    """
+    try:
+        resp = await client.get(f"{base}/api/threads/{thread_id}/state", headers=headers)
+    except Exception as exc:
+        logger.warning("intent: get thread state failed (%s) — falling back to soul-only flow", exc)
+        return []
+    if resp.status_code != 200:
+        logger.warning(
+            "intent: get thread state %s returned %s — falling back to soul-only flow",
+            thread_id,
+            resp.status_code,
+        )
+        return []
+    try:
+        payload = resp.json()
+    except Exception:
+        return []
+    values = payload.get("values") if isinstance(payload, dict) else None
+    if not isinstance(values, dict):
+        return []
+    messages = values.get("messages")
+    return messages if isinstance(messages, list) else []
+
+
+# 硬上限:超过这个轮数,无论模型怎么想,都强制移除 ask_clarification 工具。
+# 与 SOUL.md §0 「硬规则 2」对齐——"绝不存在第 4 轮"。
+_MAX_CLARIFICATION_ROUNDS = 3
+
+
 def _is_repeated_clarification(
     messages: list[dict[str, Any]],
     *,
@@ -279,33 +343,70 @@ async def stream_intent(request: Request, payload: dict = Body(...)) -> Streamin
         # 因此既无须 trust_env=False 防代理，也不存在 loopback 40s 超时这种坑。
         async with _make_loopback_client(request) as client:
             try:
+                # —— 3 轮硬上限兜底 ——
+                # 在发起 run 之前先读一次 thread state,数当前 ask_clarification 已发出次数:
+                #   past=0/1 → 正常,不动 message,不动工具白名单
+                #   past=2(本轮即将变成第 3 轮)→ 给消息加前缀,强提示"必须 INTENT_READY"
+                #   past≥3 → 直接把 ask_clarification 也加进 excluded_tools,工具物理消失
+                # 这一步是个"刹车":SOUL 让模型自律,这里让系统兜底。
+                past_clarifications = _count_past_clarifications(
+                    await _fetch_thread_messages(client, base, headers, thread_id)
+                )
+
+                # 1. ask_clarification 的工具白名单——超过硬上限就移除。
+                base_excluded_tools = [
+                    "web_search",
+                    "present_files",
+                    "view_image",
+                    "agent_orchestration",
+                    "bash",
+                    "ls",
+                    "read_file",
+                    "write_file",
+                    "str_replace",
+                    "memory",
+                    "hindsight_recall",
+                    "hindsight_reflect",
+                    "hindsight_retain",
+                    "task",
+                    "write_todos",
+                ]
+                if past_clarifications >= _MAX_CLARIFICATION_ROUNDS:
+                    base_excluded_tools.append("ask_clarification")
+                    logger.info(
+                        "intent stream: past_clarifications=%d >= %d, force-disabling "
+                        "ask_clarification to break the loop",
+                        past_clarifications,
+                        _MAX_CLARIFICATION_ROUNDS,
+                    )
+
+                # 2. message 前缀——给模型一个明确的"本轮必须收敛"信号。
+                effective_message = message
+                if past_clarifications >= _MAX_CLARIFICATION_ROUNDS:
+                    effective_message = (
+                        f"[系统强制] 已澄清 {past_clarifications} 轮,触发硬上限。"
+                        f"本轮请直接输出 [INTENT_READY],缺的字段用 assumptions 兜底,"
+                        f"绝对不要再调任何工具,也不要再询问任何问题。\n\n"
+                        f"用户最新输入:{message}"
+                    )
+                elif past_clarifications >= _MAX_CLARIFICATION_ROUNDS - 1:
+                    effective_message = (
+                        f"[系统提示] 已澄清 {past_clarifications} 轮,这是最后一轮。"
+                        f"本轮请直接输出 [INTENT_READY],未明确的字段用 assumptions 兜底。\n\n"
+                        f"用户最新输入:{message}"
+                    )
+
                 body = _run_payload(
                     agent_name=INTENT_AGENT_ID,
                     model_name=model_name,
                     thread_id=thread_id,
-                    new_message=message,
-                    # Intent agent only ever needs ask_clarification. Block
+                    new_message=effective_message,
+                    # Intent agent only ever needs ask_clarification (and even that
+                    # is yanked once past_clarifications hits the cap above). Block
                     # everything else so the model can't drift into solutioning
-                    # (e.g. saving a design doc with write_file) — the model
-                    # tends to ignore SOUL prose, hard exclusion is reliable.
-                    # 仅保留 ask_clarification；其余工具见模块 docstring「工具白名单」。
-                    excluded_tools=[
-                        "web_search",
-                        "present_files",
-                        "view_image",
-                        "agent_orchestration",
-                        "bash",
-                        "ls",
-                        "read_file",
-                        "write_file",
-                        "str_replace",
-                        "memory",
-                        "hindsight_recall",
-                        "hindsight_reflect",
-                        "hindsight_retain",
-                        "task",
-                        "write_todos",
-                    ],
+                    # (e.g. saving a design doc with write_file) — SOUL prose is
+                    # not reliable on its own, hard exclusion is.
+                    excluded_tools=base_excluded_tools,
                     skill_stop_names=[],
                 )
 
@@ -338,7 +439,13 @@ async def stream_intent(request: Request, payload: dict = Body(...)) -> Streamin
                         except json.JSONDecodeError as exc:
                             logger.warning("intent summary JSON parse failed: %s", exc)
                             summary = None
-                        ready_text = msg_text
+                        # SOUL §0 硬规则 3 要求 ```json``` 反引号闭合后不得再输出任何字符,
+                        # 但模型仍偶尔会在后面接 "那么我帮你列一下大纲……" 之类的尾巴。
+                        # 终态帧只回到 marker 块结尾 + 一个换行,把废话从 content 里裁掉——
+                        # 草稿持久化 / 跨设备重放 / 前端 splitIntentReadyBlock 的 after
+                        # 段都因此干净。流式期间用户仍会逐 token 看到那段废话,前端
+                        # Step1Panel 已经把 `parts.after` 隐藏掉,体验闭环。
+                        ready_text = msg_text[: m.end()].rstrip() + "\n"
                         break
                 if summary and ready_text:
                     final_frame = {