feat(ai写作): 素材收集专家按真实智能体运行时跑技能 + 步骤条对齐

- 新增 run_skill_via_real_agent：复用 SubagentExecutor（与 lead agent / 子智能体
  同一套运行时），加载技能 SKILL.md + 全套工具 + 懒加载沙箱，让技能的真实检索逻辑
  （读文件 / 调内网接口 / MCP）真正执行，再按 JSON 契约收割结构化素材；researcher 的
  _execute_skill 在 researcher_skill_agent 开启时切到它，失败/空结果回退直连检索。
  解决内网技能（在别的智能体里好使）在 AI 写作里检不出东西的问题。
- 步骤条轻量对齐别的智能体：技能调用拆成「读取 SKILL.md → 执行检索 → 检索到 N 条」
  三段（前两段新增，结果段沿用 skill_result）；前端 SkillStepsCard 零改动。
- 含 roundtable-planning（业务链选择 / Step2 编排）的进行中改动。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

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

@@ -140,7 +140,7 @@ export function BusinessChainPicker({
     const hasCoord = (plan.coordinatorPrompt?.trim() ?? "") !== "";
     return hasParallel || hasGoal || hasCoord ? plan : null;
   }, [selectedChain, survivingSeats]);
-  const [dagExecMode, setDagExecMode] = useState<"fast" | "sandbox">("fast");
+  const [dagExecMode, setDagExecMode] = useState<"fast" | "sandbox">("sandbox");
 
   const handleUse = () => {
     if (!canUse || !selectedChain) return;

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

@@ -460,7 +460,7 @@ export function useStep2Orchestration(
     gatherFirst = false,
     orchestrationMode,
     orchestrationPlan,
-    dagExecMode = "fast",
+    dagExecMode = "sandbox",
     buildInitialIntent,
     triggerToast,
     setRoles,

frontend-web/src/roundtable-planning/lib/business-chain.ts

@@ -192,7 +192,7 @@ export function buildChainSeatSelection(
       const agents = plan.stages
         .flatMap((st) => st.agentIds)
         .map((id) => ({ agent_id: id, name: nameOf(id), model: modelOf(id), seat_skill_directive: skillOf(id), seatMode: seatModeOf(id) }));
-      return { agents, mode: "dag", chain: chainMeta, plan, dagExecMode: "fast", seatMode: chain.seatMode, gatherFirst: chain.gatherFirst ?? false };
+      return { agents, mode: "dag", chain: chainMeta, plan, dagExecMode: "sandbox", seatMode: chain.seatMode, gatherFirst: chain.gatherFirst ?? false };
     }
   }
 

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

@@ -360,7 +360,7 @@ export default function RoundtablePlanningPage() {
   const [orchestrationPlan, setOrchestrationPlan] = useState<OrchestrationPlan | null>(null);
   // dag 执行子模式：fast（并行双列禁写）/ sandbox（并行单列可写、沙箱跟随）。
   // 由推荐弹窗的「执行方式」选择决定（仅 dag 有意义）。
-  const [dagExecMode, setDagExecMode] = useState<"fast" | "sandbox">("fast");
+  const [dagExecMode, setDagExecMode] = useState<"fast" | "sandbox">("sandbox");
   const [step2SidebarExpanded, setStep2SidebarExpanded] = useState(true);
   const [step2Runs, setStep2Runs] = useState<DraftStep2RunSnapshot[]>([]);
   const [activeStep2RunId, setActiveStep2RunId] = useState<string | null>(null);
@@ -684,7 +684,7 @@ export default function RoundtablePlanningPage() {
       setOrchestrationMode(run.orchestrationMode ?? "recommend");
       setChainMeta(run.chain ?? null);
       setOrchestrationPlan(run.orchestrationPlan ?? null);
-      setDagExecMode(run.dagExecMode ?? "fast");
+      setDagExecMode(run.dagExecMode ?? "sandbox");
       setStep3Report(run.step3 ?? null);
       step2.hydrateFromDraft(run as unknown as Step2Snapshot);
       markRestoredConsensusAsEmitted(Boolean(run.hasConsensus));
@@ -791,7 +791,7 @@ export default function RoundtablePlanningPage() {
       setOrchestrationMode(snapshot.orchestrationMode ?? "recommend");
       setChainMeta(snapshot.chain ?? null);
       setOrchestrationPlan(snapshot.orchestrationPlan ?? null);
-      setDagExecMode(snapshot.dagExecMode ?? "fast");
+      setDagExecMode(snapshot.dagExecMode ?? "sandbox");
       step2.hydrateFromDraft(snapshot as unknown as Step2Snapshot);
       markRestoredConsensusAsEmitted(Boolean(snapshot.hasConsensus));
     },
@@ -945,7 +945,7 @@ export default function RoundtablePlanningPage() {
         // chain / dag 都来自业务链条，保留来源链名；recommend 无。
         chain: result.mode === "recommend" ? null : result.chain ?? null,
         orchestrationPlan: result.mode === "dag" ? result.plan ?? null : null,
-        dagExecMode: result.mode === "dag" ? result.dagExecMode ?? "fast" : undefined,
+        dagExecMode: result.mode === "dag" ? result.dagExecMode ?? "sandbox" : undefined,
         step3: null,
       };
       setStep2Runs([...syncedRuns, newRun]);
@@ -998,7 +998,7 @@ export default function RoundtablePlanningPage() {
       setGatherFirst(result.gatherFirst ?? false);
       setChainMeta(result.mode === "recommend" ? null : result.chain ?? null);
       setOrchestrationPlan(result.mode === "dag" ? result.plan ?? null : null);
-      setDagExecMode(result.mode === "dag" ? result.dagExecMode ?? "fast" : "fast");
+      setDagExecMode(result.mode === "dag" ? result.dagExecMode ?? "sandbox" : "sandbox");
       setIsRecommendDialogOpen(false);
       setCurrentStep(2);
       triggerToast(
@@ -1527,7 +1527,7 @@ export default function RoundtablePlanningPage() {
     setOrchestrationMode("recommend");
     setChainMeta(null);
     setOrchestrationPlan(null);
-    setDagExecMode("fast");
+    setDagExecMode("sandbox");
     setStep3Report(null);
     // 停掉页面层的 Step3 三个生成流（新任务不该带着旧任务的流跑）+ 复位进入闸门，
     // 并停止后台作业实时跟随（作业本身继续在后台跑）。