feat(task): 添加任务修订功能和超时处理机制

添加了 `revise_task` 路由动作类型，允许用户修改、纠正或重新执行最新活动任务结果。
实现了工具失败指导原则，防止相同类别工具重复失败。
为任务规划器添加了超时处理机制，避免长时间等待。

BREAKING CHANGE: 任务路由逻辑已更新，新增 `revise_task` 动作类型。

fix(api): 修复任务详情API返回完整流程投影

修复了任务详情API端点，现在会包含过滤后的流程运行、事件和工件信息，
并确保时间戳字段正确序列化。

refactor(engine): 优化任务技能解析器摘要节点处理

改进了任务技能解析器对摘要节点的处理逻辑，对于仅依赖文本生成功能的摘要节
点不再分配具体技能，直接使用依赖项输出进行汇总。

test: 增加任务修订和超时处理测试用例

添加了测试用例验证任务修订输入记录反馈、超时回退到单模式以及
摘要节点技能解析等新功能。

app-instance/backend/beaver/services/agent_service.py

@@ -581,8 +581,96 @@ class AgentService:
             if active_task is None or decision.starts_new_task
             else active_task
         )
+        if active_task is not None and decision.action == "revise_task" and task.task_id == active_task.task_id:
+            task = self._record_revision_feedback_for_task(
+                loaded,
+                task=task,
+                session_id=session_id,
+                comment=message,
+            )
         return await self._run_task_mode(message, runner=runner, kwargs=kwargs, task=task)
 
+    def _record_revision_feedback_for_task(
+        self,
+        loaded: Any,
+        *,
+        task: TaskRecord,
+        session_id: str,
+        comment: str,
+    ) -> TaskRecord:
+        """Mark the latest feedback-eligible run as revised before continuing a task."""
+
+        if task.status not in {"awaiting_feedback", "needs_revision"}:
+            return task
+        run_id = next((item for item in reversed(task.run_ids) if item), None)
+        if not run_id:
+            return task
+
+        existing = next((item for item in task.feedback if item.get("run_id") == run_id), None)
+        if existing is not None:
+            if existing.get("feedback_type") != "revise":
+                return task
+            updated = task
+            already_recorded = True
+        else:
+            task_service = self._require_loaded(loaded, "task_service")
+            updated = task_service.add_feedback(
+                task.task_id,
+                feedback_type="revise",
+                comment=comment,
+                run_id=run_id,
+            )
+            already_recorded = False
+
+        session_manager = self._require_loaded(loaded, "session_manager")
+        session_manager.update_latest_assistant_event_payload(
+            session_id,
+            run_id,
+            {
+                "task_id": updated.task_id,
+                "task_status": updated.status,
+                "feedback_state": "revise",
+            },
+        )
+        if already_recorded:
+            return updated
+
+        session_manager.append_message(
+            session_id,
+            run_id=run_id,
+            role="system",
+            event_type="task_feedback_recorded",
+            event_payload={
+                "task_id": updated.task_id,
+                "feedback_type": "revise",
+                "comment": comment,
+                "task_status": updated.status,
+                "auto_recorded": True,
+            },
+            content=comment,
+            context_visible=False,
+        )
+        validation = ValidationResult.from_dict(updated.validation_result)
+        run_memory_store = self._require_loaded(loaded, "run_memory_store")
+        run_memory_store.update_run_record(
+            run_id,
+            success=False,
+            feedback={
+                "feedback_type": "revise",
+                "comment": comment,
+                "task_status": updated.status,
+            },
+        )
+        run_memory_store.update_skill_effects_for_run(
+            run_id,
+            success=False,
+            feedback_score=self._feedback_score_for_learning("revise", validation),
+            notes=comment.strip() or "revise",
+        )
+        skill_learning_service = self._require_loaded(loaded, "skill_learning_service")
+        skill_learning_service.rescore_skill_versions()
+        return updated
+
     async def _run_task_mode(
         self,
         message: str,
@@ -1018,7 +1106,11 @@ class AgentService:
                     if plan.final_synthesis_instruction
                     else None
                 ),
-                "Use the team outputs as internal evidence. Produce the final user-facing answer yourself.",
+                (
+                    "Use successful team outputs as internal evidence. If one or more nodes failed, "
+                    "do not blindly repeat failed tool calls. Produce a user-visible fallback answer "
+                    "with available evidence and clearly state any missing or uncertain data."
+                ),
             ]
             if item
         )
@@ -1031,7 +1123,11 @@ class AgentService:
                 f"Planner reason: {plan.reason}",
                 f"Strategy: {plan.graph.strategy if plan.graph else ''}",
                 f"Error: {error}",
-                "Proceed as the main agent and produce the best possible final answer.",
+                (
+                    "Proceed as the main agent. Do not blindly repeat failed tool calls; "
+                    "produce a user-visible fallback answer with available evidence and clearly "
+                    "state any missing or uncertain data."
+                ),
             ]
         )