merge iteration loop p2 recommendation quality

# Conflicts:
#	docs/PMFLOW_PRODUCT_DELIVERY_STANDARD.md

CLAUDE.md

@@ -113,8 +113,8 @@ frontend/
   - `frontend/src/features/personal-space/`（个人空间首页；通过 `frontend/src/api/space.ts` 只读聚合项目摘要、PMFlow OS truth-state / execution-summary，承载继续推进、待确认、AI 正在推进、草案箱和最近项目入口）
   - `frontend/src/features/project-launch/`（首次进入页 / 新项目启动台；默认调用真实 `frontend/src/api/pmflowLaunch.ts` draft/confirm API 创建项目，只有显式 `?demo=1` 或 `?dataSource=demo` 才启用本地 mock 演示状态机）
   - `frontend/src/features/account-settings/`（`/me` 个人中心与设置；通过 `frontend/src/api/accountSettings.ts` 管理个人资料、默认语言、AI 回答深度、证据展示偏好和通知偏好，移动端使用分组列表与 bottom sheet，不进入 auth core）
-  - `frontend/src/features/pmflow-os/`（PMFlow OS 前端产品域；共享 truth store 已支持 `?projectId=<id>` 读取真实后端 projection，真实模式下 message、`proposed_items` extraction、ChangeItem `accept/edit/defer`、ChangeSet `commit` 与 iteration recommendation -> ChangeSet draft 调后端，显式 `?demo=1` 才启用本地演示状态机）
+  - `frontend/src/features/pmflow-os/`（PMFlow OS 前端产品域；共享 truth store 已支持 `?projectId=<id>` 读取真实后端 projection，真实模式下 message、`proposed_items` extraction、ChangeItem `accept/edit/defer`、ChangeSet `commit`、单条 iteration recommendation -> ChangeSet draft 与同组 iteration recommendation 批量 -> ChangeSet draft 均调后端，显式 `?demo=1` 才启用本地演示状态机）
-  - `backend/app/api/pmflow_os_api/` + `backend/app/services/pmflow_os_projection_service.py` + `backend/app/services/pmflow_change_review_service.py` + `backend/app/schemas/pmflow_os.py` + `backend/app/models/pmflow_change_review.py`（PMFlow OS A3/C1/C2 真实路径；提供 `truth-state`、`execution-summary`、thread message、`proposed_items` extraction、ChangeSet detail、ChangeItem `accept/edit/defer`、ChangeSet `commit`、execution handoff/progress writeback、release validation run、release feedback ingest 与 recommendation change-set draft；复用 `truth_nodes / tasks / semantic_facts / ai_project_control / dev_sync / FeedbackIngest / IterationRecommendation` 并投影 `pmflow_threads / pmflow_thread_messages / pmflow_change_sets / pmflow_change_items / pmflow_approvals / pmflow_evidence_links`；C2 commit 默认 atomic，`ModelDecision / MetricSignal` 暂 fail-closed）
+  - `backend/app/api/pmflow_os_api/` + `backend/app/services/pmflow_os_projection_service.py` + `backend/app/services/pmflow_change_review_service.py` + `backend/app/schemas/pmflow_os.py` + `backend/app/models/pmflow_change_review.py`（PMFlow OS A3/C1/C2 真实路径；提供 `truth-state`、`execution-summary`、thread message、`proposed_items` extraction、ChangeSet detail、ChangeItem `accept/edit/defer`、ChangeSet `commit`、execution handoff/progress writeback、release validation run、release feedback ingest、recommendation change-set draft 与 recommendation group change-set draft；复用 `truth_nodes / tasks / semantic_facts / ai_project_control / dev_sync / FeedbackIngest / IterationRecommendation` 并投影 `pmflow_threads / pmflow_thread_messages / pmflow_change_sets / pmflow_change_items / pmflow_approvals / pmflow_evidence_links`；C2 commit 默认 atomic，`ModelDecision / MetricSignal` 暂 fail-closed）
   - `backend/app/api/pmflow_launch_api/` + `backend/app/services/pmflow_launch_service.py` + `backend/app/schemas/pmflow_launch.py` + `backend/app/models/pmflow_launch.py`（Project Launch Phase B 真实创建链；`drafts` 只写 `pmflow_launch_drafts`，`confirm` 在同一事务写真实 Project、启动线程、启动消息、初始 truth_nodes、tasks 与 semantic_facts seed）
   - `docs/architecture/PMFLOW_OS_PRODUCT_CONTRACT.md`（PMFlow OS 产品合同真源；真实后端接入前先读，锁定本地 truth store 对象与 `truth_nodes / tasks / semantic_facts / ai_project_control / dev_sync` 的映射；Phase C1 review 表和审查 API 已实现，Phase C2 backend/frontend commit 已实现）
   - `frontend/src/pages/inbox/`（消息台）
@@ -134,7 +134,7 @@ frontend/
 - `frontend/src/features/project-launch/` 是新项目启动台独立产品域：`types.ts / useProjectLaunchState.ts` 承载状态合同和真实 API 编排，`mockData.ts` 只在显式 demo 模式下使用；默认路径调用 `POST /pmflow-launch/drafts` 生成可审查草案，确认后调用 `POST /pmflow-launch/drafts/{draft_id}/confirm` 创建真实项目并跳转 `/pmflow-os?projectId=<id>`。`LaunchShell` 桌面端复用 `frontend/src/features/personal-space/components/SpaceChrome.tsx` 的空间左侧栏和顶部条，主画布仍由 `IntentInputPanel / AIProgressCanvas / HumanReviewCanvas / DraftEditCanvas / FailureStatePanel / SuccessTransitionPanel` 承接，移动端仍由 `MobileLaunchLayout` 承接；`frontend/src/pages/ProjectLaunchPage.tsx` 仅做薄入口。
 - `frontend/src/features/personal-space/` 是空间层入口：`PersonalSpacePage.tsx` 承接桌面个人工作面与独立移动单手布局，`frontend/src/api/space.ts` 只读聚合 `projects` summary、PMFlow OS `truth-state`、`execution-summary` 并生成继续推进、待确认、发布关注、发布后反馈关注、下一轮迭代草稿关注、AI 正在推进、草案箱、最近项目摘要；Space P3 的发布关注优先消费 `truth-state.releases` 的 release validation 投影，没有验证投影时才回落 `execution-summary.releaseGate`；Space P4/P5 的发布后反馈/下一轮草稿关注只读消费 `execution-summary.releaseSignals`，点击只跳 `/release-growth?projectId=<id>` 或 `/change-review?projectId=<id>&changeSetId=<id>`；不在 Space 写 Project Memory、ProgressEvent、ReleaseValidationRun、FeedbackIngest、RevenueEvent、ComplianceCheckRun、ChangeSet 状态或发布执行。`components/SpaceChrome.tsx` 承接空间层左侧栏、顶部条与空间用量组件，并被 `/project-launch` 桌面首次新建壳复用；`frontend/src/pages/PersonalSpacePage.tsx` 仅做薄入口。
 - `frontend/src/features/account-settings/` 是 Account Settings 独立产品域：`AccountSettingsPanel.tsx` 承接 `/me` 桌面设置页、移动端分组列表与 bottom sheet，`frontend/src/api/accountSettings.ts` 定义个人资料、默认语言、AI 回答深度、证据展示偏好和四类通知偏好合同；当前已有账号同步 API fallback、本地 UI 通知消费模型、Release/Growth 偏好 helper 和 post-release / next change review 偏好 helper，`releaseGate / failureAlert / pendingConfirmation` 只影响页面内提醒表达，不代表真实邮件、桌面推送或系统通知中心已接入，不改 OAuth/JWT/auth core；`frontend/src/features/pmflow-os/pages/AccountPage.tsx` 仅做薄入口并复用既有退出登录 bottom sheet。
-- `frontend/src/features/pmflow-os/` 当前承载新的对话驱动前端产品域：`store.tsx` 提供共享 truth store；真实模式通过 `?projectId=<id>` 调用 `frontend/src/api/pmflowOs.ts` 读取后端 `ProjectTruthState` projection，并调用真实 message、`proposed_items` extraction、ChangeItem `accept/edit/defer`、ChangeSet `commit`、Execution Handoff create/consume、execution progress writeback、release validation run create、release feedback ingest 与 iteration recommendation -> ChangeSet draft API，成功后 refetch truth-state / execution-summary；显式 `?demo=1` 才启用 `command_idle -> ai_extracting -> extraction_ready -> review_open -> writeback_success` 本地演示状态机，以及本地 `promoteAcceptedToGraph` 等演示动作。`pages/` 下包含 `CommandCenter / StructuredExtraction / ChangeReview / MemoryWritebackResult / WorkGraph / ExecutionBridge / ReleaseGrowth / TeamLayer / ProjectMemory / Account` 主页面；`ExecutionBridgePage` 已支持创建 handoff、运行端领取确认和执行进度回写；`ReleaseGrowthPage` 真实模式展示 `execution-summary.releaseGate`、execution evidence、阻塞原因、下一步动作、`latestValidationRun` 和 `releaseSignals`，可创建 release validation run、摄取发布后 feedback、从真实 recommendation 生成下一轮待确认 ChangeSet draft，并消费 Account notification preference 静默强调型提醒，但不隐藏真实证据、审计状态或发布后信号；静态发布/增长原型仅保留在显式 demo 模式；后端只记录 progress/release gate/release validation/release feedback/recommendation draft 元数据，不执行本机命令、真实发布命令、外部支付、真实推送、自动 Project Memory commit 或正式任务创建；`ProjectMemoryPage` 在真实模式消费后端 `truth_nodes + semantic_facts` 的 `projectMemory` 投影，真实模式下 accepted/edited ChangeItem 只有 commit 成功后才进入 Project Memory。`components/` 下除了 `ProjectStageRail / ThreadComposer / ExtractionRail / ChangeGroupPanel / ObjectGraphSection / ContextPacketPanel / ExecutionTimeline / ReleaseGatePanel / TeamFeedPanel` 之外，新增 `ProjectStructuredLoop / PMFlowOsFrame / AccountControls / FeedbackPrimitives` 等行为组件，用来承接项目指挥台、结构化提取、待确认审查、项目记忆写入结果、项目内面包屑、项目切换器、账号退出登录交互，以及 `PageBanner / InlineNotice / HumanDecisionBar / ConfirmDialog / MobileBottomSheet` 这组契约化提示提醒组件；`components/MobileBottomNav.tsx` 是正式应用壳唯一移动底部导航，固定为 `空间 / 指挥台 / 待确认 / 动态 / 我的`。
+- `frontend/src/features/pmflow-os/` 当前承载新的对话驱动前端产品域：`store.tsx` 提供共享 truth store；真实模式通过 `?projectId=<id>` 调用 `frontend/src/api/pmflowOs.ts` 读取后端 `ProjectTruthState` projection，并调用真实 message、`proposed_items` extraction、ChangeItem `accept/edit/defer`、ChangeSet `commit`、Execution Handoff create/consume、execution progress writeback、release validation run create、release feedback ingest、单条 iteration recommendation -> ChangeSet draft 与同组 iteration recommendation 批量 -> ChangeSet draft API，成功后 refetch truth-state / execution-summary；显式 `?demo=1` 才启用 `command_idle -> ai_extracting -> extraction_ready -> review_open -> writeback_success` 本地演示状态机，以及本地 `promoteAcceptedToGraph` 等演示动作。`pages/` 下包含 `CommandCenter / StructuredExtraction / ChangeReview / MemoryWritebackResult / WorkGraph / ExecutionBridge / ReleaseGrowth / TeamLayer / ProjectMemory / Account` 主页面；`ExecutionBridgePage` 已支持创建 handoff、运行端领取确认和执行进度回写；`ReleaseGrowthPage` 真实模式展示 `execution-summary.releaseGate`、execution evidence、阻塞原因、下一步动作、`latestValidationRun` 和 `releaseSignals`，可创建 release validation run、摄取发布后 feedback、展示 recommendation quality grouping、从真实 recommendation 单条或同组批量生成下一轮待确认 ChangeSet draft，并消费 Account notification preference 静默强调型提醒，但不隐藏真实证据、审计状态或发布后信号；`ChangeReviewPage` 会展示来自 release feedback recommendation group 的组、反馈、风险、影响和审查优先级，帮助 reviewer 判断为什么生成这一组待确认变更；静态发布/增长原型仅保留在显式 demo 模式；后端只记录 progress/release gate/release validation/release feedback/recommendation draft 元数据，不执行本机命令、真实发布命令、外部支付、真实推送、自动 Project Memory commit 或正式任务创建；`ProjectMemoryPage` 在真实模式消费后端 `truth_nodes + semantic_facts` 的 `projectMemory` 投影，真实模式下 accepted/edited ChangeItem 只有 commit 成功后才进入 Project Memory。`components/` 下除了 `ProjectStageRail / ThreadComposer / ExtractionRail / ChangeGroupPanel / ObjectGraphSection / ContextPacketPanel / ExecutionTimeline / ReleaseGatePanel / TeamFeedPanel` 之外，新增 `ProjectStructuredLoop / PMFlowOsFrame / AccountControls / FeedbackPrimitives` 等行为组件，用来承接项目指挥台、结构化提取、待确认审查、项目记忆写入结果、项目内面包屑、项目切换器、账号退出登录交互，以及 `PageBanner / InlineNotice / HumanDecisionBar / ConfirmDialog / MobileBottomSheet` 这组契约化提示提醒组件；`components/MobileBottomNav.tsx` 是正式应用壳唯一移动底部导航，固定为 `空间 / 指挥台 / 待确认 / 动态 / 我的`。
 - 旧登录实现 `frontend/src/components/ui/sign-in.tsx` 与 `frontend/src/components/ui/demo.tsx` 已移除，`frontend/src/components/login/` 是当前唯一登录真源。
 
 ### 5.1 `frontend/src/components/sstc/` 目录说明（模块化拆分后）

backend/app/api/pmflow_os_api/__init__.py

@@ -18,6 +18,7 @@ from ...schemas.pmflow_os import (
     PMFlowExecutionProgressEventRequest,
     PMFlowExtractionCreateRequest,
     PMFlowIterationRecommendationChangeSetDraftRequest,
+    PMFlowIterationRecommendationGroupChangeSetDraftRequest,
     PMFlowReleaseFeedbackIngestRequest,
     PMFlowReleaseValidationRunCreateRequest,
     PMFlowThreadMessageCreateRequest,
@@ -419,6 +420,41 @@ async def create_pmflow_os_iteration_recommendation_change_set_draft(
     return success_response(payload)
 
 
+@router.post("/projects/{project_id}/iteration-recommendation-groups/{group_key}/change-set-drafts")
+async def create_pmflow_os_iteration_recommendation_group_change_set_draft(
+    project_id: int,
+    group_key: str,
+    body: PMFlowIterationRecommendationGroupChangeSetDraftRequest | None = None,
+    db: AsyncSession = Depends(get_db),
+    current_user: User = Depends(get_current_user),
+):
+    tenant_id = int(current_user.tenant_id)
+    await _ensure_project_member_or_manager(
+        db,
+        tenant_id=tenant_id,
+        project_id=int(project_id),
+        current_user=current_user,
+    )
+    service = PMFlowOSProjectionService(db, tenant_id=tenant_id)
+    try:
+        payload = await service.create_iteration_recommendation_group_change_set_draft(
+            project_id=int(project_id),
+            group_key=group_key,
+            body=body or PMFlowIterationRecommendationGroupChangeSetDraftRequest(),
+            current_user_id=int(current_user.id),
+        )
+    except ValueError as exc:
+        code = str(exc)
+        if code == "iteration_recommendation_group_not_found":
+            return JSONResponse(status_code=404, content=error_response(404, "iteration_recommendation_group_not_found"))
+        if code == "iteration_recommendation_not_found":
+            return JSONResponse(status_code=404, content=error_response(404, "iteration_recommendation_not_found"))
+        if code == "project_not_found":
+            raise HTTPException(status_code=404, detail="project_not_found") from exc
+        raise
+    return success_response(payload)
+
+
 @router.post("/projects/{project_id}/execution-handoffs/{prompt_package_id}/consume")
 async def consume_pmflow_os_execution_handoff(
     project_id: int,

backend/app/schemas/pmflow_os.py

@@ -150,6 +150,11 @@ class PMFlowIterationRecommendationChangeSetDraftRequest(BaseModel):
     note: Optional[str] = Field(default=None, max_length=4000)
 
 
+class PMFlowIterationRecommendationGroupChangeSetDraftRequest(BaseModel):
+    recommendation_ids: list[str] = Field(default_factory=list)
+    note: Optional[str] = Field(default=None, max_length=4000)
+
+
 class PMFlowChangeSetDetailResponse(BaseModel):
     changeSet: dict[str, Any]
     items: list[dict[str, Any]] = Field(default_factory=list)

backend/app/services/pmflow_os_projection_service.py

@@ -44,6 +44,7 @@ from ..schemas.pmflow_os import (
     PMFlowExecutionContext,
     PMFlowFeedEvent,
     PMFlowIterationRecommendationChangeSetDraftRequest,
+    PMFlowIterationRecommendationGroupChangeSetDraftRequest,
     PMFlowMetricSignal,
     PMFlowModelDecision,
     PMFlowProjectMemory,
@@ -521,6 +522,168 @@ class PMFlowOSProjectionService:
             "warnings": [],
         }
 
+    async def create_iteration_recommendation_group_change_set_draft(
+        self,
+        *,
+        project_id: int,
+        group_key: str,
+        body: PMFlowIterationRecommendationGroupChangeSetDraftRequest,
+        current_user_id: Optional[int] = None,
+    ) -> dict[str, Any]:
+        project = await self._load_project(project_id=project_id)
+        rows = await self._load_iteration_recommendations(project_id=project_id)
+        feedback_by_id = await self._load_feedback_ingests_for_recommendations(project_id=project_id, recommendations=rows)
+        quality_by_id = {
+            int(row.id): self._recommendation_quality_profile(
+                recommendation=row,
+                feedback=feedback_by_id.get(int(row.feedback_ingest_id)) if row.feedback_ingest_id is not None else None,
+            )
+            for row in rows
+        }
+        requested_ids = self._parse_recommendation_ids(body.recommendation_ids)
+        normalized_group_key = self._normalize_group_key(group_key)
+        selected = [
+            row
+            for row in rows
+            if (requested_ids and int(row.id) in requested_ids) or (not requested_ids and quality_by_id[int(row.id)]["groupKey"] == normalized_group_key)
+        ]
+        if requested_ids and len({int(row.id) for row in selected}) != len(requested_ids):
+            raise ValueError("iteration_recommendation_not_found")
+        if not selected:
+            raise ValueError("iteration_recommendation_group_not_found")
+        selected_quality = self._finalize_recommendation_quality(
+            recommendations=selected,
+            quality_by_id={int(row.id): quality_by_id[int(row.id)] for row in selected},
+        )
+        canonical_group_key = str(next(iter(selected_quality.values())).get("groupKey") or normalized_group_key)
+        existing_change_set = await self._find_existing_recommendation_group_change_set(
+            project_id=project_id,
+            recommendations=selected,
+        )
+        if existing_change_set is not None:
+            await self._mark_recommendations_for_group_change_set(
+                recommendations=selected,
+                change_set_id=int(existing_change_set.id),
+                group_key=canonical_group_key,
+                current_user_id=current_user_id,
+                reused=True,
+            )
+            await self.db.commit()
+            detail = await PMFlowChangeReviewService(self.db, tenant_id=self.tenant_id).get_change_set_detail(
+                project_id=project_id,
+                change_set_id=int(existing_change_set.id),
+            )
+            updated_summary = await self.get_execution_summary(project_id=project_id, current_user_id=current_user_id)
+            return {
+                "changeSetDetail": detail,
+                "executionSummary": updated_summary.model_dump(mode="json"),
+                "warnings": ["iteration_recommendation_group_changeset_reused"],
+            }
+
+        thread = await self._resolve_or_create_iteration_thread(project=project)
+        message = PMFlowThreadMessage(
+            tenant_id=self.tenant_id,
+            thread_id=int(thread.id),
+            project_id=int(project_id),
+            role="assistant",
+            content=self._iteration_group_message(
+                recommendations=selected,
+                feedback_by_id=feedback_by_id,
+                quality_by_id=selected_quality,
+                note=body.note,
+            ),
+            source="pmflow_os",
+            attachments=[],
+            message_metadata={
+                "source": "iteration_recommendation_group",
+                "group_key": canonical_group_key,
+                "recommendation_ids": [str(row.id) for row in selected],
+                "created_by": None if current_user_id is None else int(current_user_id),
+                "auto_commit": False,
+                "note": body.note,
+            },
+            status="sent",
+        )
+        self.db.add(message)
+        await self.db.flush()
+
+        items = self._iteration_recommendation_group_change_items(
+            recommendations=selected,
+            feedback_by_id=feedback_by_id,
+            quality_by_id=selected_quality,
+            project_id=project_id,
+        )
+        change_set = PMFlowChangeSet(
+            tenant_id=self.tenant_id,
+            project_id=int(project_id),
+            thread_id=int(thread.id),
+            source_message_id=int(message.id),
+            status="open",
+            summary=self._iteration_group_change_set_summary(recommendations=selected, quality_by_id=selected_quality),
+            headline=self._iteration_group_change_set_headline(recommendations=selected, quality_by_id=selected_quality),
+            extraction_scope="iteration_recommendation_group",
+            routing_meta={
+                "source": "release_feedback_iteration_group",
+                "groupKey": canonical_group_key,
+                "recommendationIds": [str(row.id) for row in selected],
+                "releaseFeedbackLoop": True,
+                "autoCommit": False,
+                "note": body.note,
+            },
+            llm_call_id=None,
+            created_by=int(current_user_id or project.owner_id),
+        )
+        self.db.add(change_set)
+        await self.db.flush()
+
+        for item in items:
+            item.change_set_id = int(change_set.id)
+            self.db.add(item)
+        await self.db.flush()
+
+        evidence = PMFlowEvidenceLink(
+            tenant_id=self.tenant_id,
+            project_id=int(project_id),
+            change_set_id=int(change_set.id),
+            change_item_id=None,
+            source_type="iteration_recommendation_group",
+            source_id=canonical_group_key,
+            title="发布后反馈迭代建议组",
+            excerpt=self._iteration_group_excerpt(recommendations=selected, feedback_by_id=feedback_by_id),
+            display_label="Release Feedback Group",
+            evidence_metadata={
+                "groupKey": canonical_group_key,
+                "recommendationIds": [str(row.id) for row in selected],
+                "feedbackIngestIds": [
+                    str(row.feedback_ingest_id) for row in selected if row.feedback_ingest_id is not None
+                ],
+                "quality": next(iter(selected_quality.values()), {}),
+                "change_set_created": True,
+            },
+        )
+        self.db.add(evidence)
+
+        await self._mark_recommendations_for_group_change_set(
+            recommendations=selected,
+            change_set_id=int(change_set.id),
+            group_key=canonical_group_key,
+            current_user_id=current_user_id,
+            reused=False,
+        )
+        await self.db.commit()
+        await self.db.refresh(change_set)
+
+        detail = await PMFlowChangeReviewService(self.db, tenant_id=self.tenant_id).get_change_set_detail(
+            project_id=project_id,
+            change_set_id=int(change_set.id),
+        )
+        updated_summary = await self.get_execution_summary(project_id=project_id, current_user_id=current_user_id)
+        return {
+            "changeSetDetail": detail,
+            "executionSummary": updated_summary.model_dump(mode="json"),
+            "warnings": [],
+        }
+
     async def consume_execution_handoff(
         self,
         *,
@@ -705,6 +868,14 @@ class PMFlowOSProjectionService:
             raise ValueError("iteration_recommendation_not_found")
         return row
 
+    async def _load_iteration_recommendations(self, *, project_id: int) -> list[IterationRecommendation]:
+        result = await self.db.execute(
+            select(IterationRecommendation)
+            .where(and_(IterationRecommendation.tenant_id == self.tenant_id, IterationRecommendation.project_id == int(project_id)))
+            .order_by(IterationRecommendation.updated_at.desc(), IterationRecommendation.created_at.desc(), IterationRecommendation.id.desc())
+        )
+        return list(result.scalars().all())
+
     async def _load_recommendation_change_set(self, *, project_id: int, change_set_id: int) -> Optional[PMFlowChangeSet]:
         result = await self.db.execute(
             select(PMFlowChangeSet).where(
@@ -717,6 +888,24 @@ class PMFlowOSProjectionService:
         )
         return result.scalar_one_or_none()
 
+    async def _find_existing_recommendation_group_change_set(
+        self,
+        *,
+        project_id: int,
+        recommendations: list[IterationRecommendation],
+    ) -> Optional[PMFlowChangeSet]:
+        candidate_ids: list[int] = []
+        for recommendation in recommendations:
+            summary = dict(recommendation.recommendation_summary or {})
+            change_set_id = self._optional_int(summary.get("changeSetId"), error_code="iteration_recommendation_not_found")
+            if change_set_id is not None:
+                candidate_ids.append(change_set_id)
+        for change_set_id in candidate_ids:
+            row = await self._load_recommendation_change_set(project_id=project_id, change_set_id=change_set_id)
+            if row is not None:
+                return row
+        return None
+
     async def _load_feedback_ingest(self, *, project_id: int, feedback_ingest_id: Optional[int]) -> Optional[FeedbackIngest]:
         if feedback_ingest_id is None:
             return None
@@ -731,6 +920,26 @@ class PMFlowOSProjectionService:
         )
         return result.scalar_one_or_none()
 
+    async def _load_feedback_ingests_for_recommendations(
+        self,
+        *,
+        project_id: int,
+        recommendations: list[IterationRecommendation],
+    ) -> dict[int, FeedbackIngest]:
+        ids = sorted({int(row.feedback_ingest_id) for row in recommendations if row.feedback_ingest_id is not None})
+        if not ids:
+            return {}
+        result = await self.db.execute(
+            select(FeedbackIngest).where(
+                and_(
+                    FeedbackIngest.tenant_id == self.tenant_id,
+                    FeedbackIngest.project_id == int(project_id),
+                    FeedbackIngest.id.in_(ids),
+                )
+            )
+        )
+        return {int(row.id): row for row in result.scalars().all()}
+
     async def _resolve_or_create_iteration_thread(self, *, project: Project) -> PMFlowThreadModel:
         result = await self.db.execute(
             select(PMFlowThreadModel)
@@ -776,7 +985,26 @@ class PMFlowOSProjectionService:
             .limit(10)
         )
         recommendation_rows = list(recommendation_result.scalars().all())
-        recommendations = [self._serialize_iteration_recommendation(item) for item in recommendation_rows]
+        feedback_by_id_for_recommendations = await self._load_feedback_ingests_for_recommendations(
+            project_id=project_id,
+            recommendations=recommendation_rows,
+        )
+        base_quality_by_id = {
+            int(item.id): self._recommendation_quality_profile(
+                recommendation=item,
+                feedback=feedback_by_id_for_recommendations.get(int(item.feedback_ingest_id)) if item.feedback_ingest_id is not None else None,
+            )
+            for item in recommendation_rows
+        }
+        quality_by_id = self._finalize_recommendation_quality(
+            recommendations=recommendation_rows,
+            quality_by_id=base_quality_by_id,
+        )
+        recommendations = [
+            self._serialize_iteration_recommendation(item, quality=quality_by_id.get(int(item.id)))
+            for item in recommendation_rows
+        ]
+        recommendation_groups = self._serialize_recommendation_groups(recommendations)
 
         revenue_result = await self.db.execute(
             select(RevenueEvent)
@@ -818,6 +1046,7 @@ class PMFlowOSProjectionService:
             "feedback": feedback_items,
             "revenueEvents": revenue_events,
             "recommendations": recommendations,
+            "recommendationGroups": recommendation_groups,
             "nextActions": self._release_signal_next_actions(
                 feedback_items=feedback_items,
                 recommendations=recommendations,
@@ -1882,10 +2111,16 @@ class PMFlowOSProjectionService:
             "updatedAt": ingest.updated_at.isoformat() if ingest.updated_at else None,
         }
 
-    def _serialize_iteration_recommendation(self, recommendation: IterationRecommendation) -> dict[str, Any]:
+    def _serialize_iteration_recommendation(
+        self,
+        recommendation: IterationRecommendation,
+        *,
+        quality: Optional[dict[str, Any]] = None,
+    ) -> dict[str, Any]:
         summary = dict(recommendation.recommendation_summary or {})
         change_set_id = summary.get("changeSetId")
         normalized_change_set_id = None if change_set_id is None or str(change_set_id).strip() == "" else str(change_set_id)
+        quality_payload = dict(quality or {})
         return {
             "id": str(recommendation.id),
             "projectId": str(recommendation.project_id),
@@ -1898,10 +2133,141 @@ class PMFlowOSProjectionService:
             "proposedTasks": list(recommendation.proposed_tasks or []),
             "changeSetCreated": bool(summary.get("changeSetCreated") or normalized_change_set_id),
             "changeSetId": normalized_change_set_id,
+            "groupKey": quality_payload.get("groupKey") or summary.get("groupKey"),
+            "duplicateOf": quality_payload.get("duplicateOf"),
+            "mergedCount": int(quality_payload.get("mergedCount") or 1),
+            "impactLevel": quality_payload.get("impactLevel") or "medium",
+            "riskLevel": quality_payload.get("riskLevel") or "medium",
+            "suggestedChangeTypes": list(quality_payload.get("suggestedChangeTypes") or ["Requirement", "Decision"]),
+            "reviewPriority": quality_payload.get("reviewPriority") or "p2",
+            "quality": quality_payload,
             "createdAt": recommendation.created_at.isoformat() if recommendation.created_at else None,
             "updatedAt": recommendation.updated_at.isoformat() if recommendation.updated_at else None,
         }
 
+    def _recommendation_quality_profile(
+        self,
+        *,
+        recommendation: IterationRecommendation,
+        feedback: Optional[FeedbackIngest],
+    ) -> dict[str, Any]:
+        summary = dict(recommendation.recommendation_summary or {})
+        patch = dict(recommendation.proposed_charter_patch or {})
+        focus_areas = patch.get("focus_areas") or patch.get("focusAreas") or summary.get("focus_areas") or summary.get("focusAreas") or []
+        if not isinstance(focus_areas, list):
+            focus_areas = [str(focus_areas)]
+        focus_areas = [str(item).strip() for item in focus_areas if str(item).strip()]
+        group_seed = focus_areas[0] if focus_areas else (feedback.title if feedback is not None and feedback.title else f"recommendation-{recommendation.id}")
+        group_key = self._normalize_group_key(group_seed)
+        priority = str(recommendation.priority or "medium").lower()
+        proposed_tasks = [item for item in list(recommendation.proposed_tasks or []) if isinstance(item, dict)]
+        metrics = dict(feedback.metrics or {}) if feedback is not None else {}
+        negative_feedback = self._safe_number(metrics.get("negative_feedback_count") or metrics.get("negativeFeedbackCount"))
+        suggested_change_types = ["Requirement", "Decision"]
+        if proposed_tasks:
+            suggested_change_types.append("Task")
+        impact_level = "high" if priority in {"high", "critical"} or negative_feedback >= 3 or len(proposed_tasks) >= 3 else "medium"
+        if negative_feedback <= 0 and not proposed_tasks and priority == "low":
+            impact_level = "low"
+        risk_level = "high" if priority in {"high", "critical"} else ("medium" if proposed_tasks or impact_level == "medium" else "low")
+        if risk_level == "high":
+            suggested_change_types.append("Risk")
+        review_priority = "p0" if risk_level == "high" and impact_level == "high" else ("p1" if impact_level == "high" or risk_level == "medium" else "p2")
+        return {
+            "groupKey": group_key,
+            "groupLabel": group_seed,
+            "duplicateOf": None,
+            "mergedCount": 1,
+            "impactLevel": impact_level,
+            "riskLevel": risk_level,
+            "suggestedChangeTypes": self._unique_strings(suggested_change_types),
+            "reviewPriority": review_priority,
+            "rationale": self._recommendation_quality_rationale(
+                focus_areas=focus_areas,
+                negative_feedback=negative_feedback,
+                task_count=len(proposed_tasks),
+                priority=priority,
+            ),
+        }
+
+    def _finalize_recommendation_quality(
+        self,
+        *,
+        recommendations: list[IterationRecommendation],
+        quality_by_id: dict[int, dict[str, Any]],
+    ) -> dict[int, dict[str, Any]]:
+        groups: dict[str, list[int]] = {}
+        for recommendation in recommendations:
+            quality = quality_by_id.get(int(recommendation.id)) or {}
+            groups.setdefault(str(quality.get("groupKey") or f"recommendation-{recommendation.id}"), []).append(int(recommendation.id))
+        finalized: dict[int, dict[str, Any]] = {}
+        for group_key, ids in groups.items():
+            first_id = str(ids[0])
+            merged_count = len(ids)
+            for rec_id in ids:
+                quality = dict(quality_by_id.get(rec_id) or {})
+                quality["groupKey"] = group_key
+                quality["mergedCount"] = merged_count
+                quality["duplicateOf"] = None if str(rec_id) == first_id else first_id
+                finalized[rec_id] = quality
+        return finalized
+
+    def _serialize_recommendation_groups(self, recommendations: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        groups: dict[str, list[dict[str, Any]]] = {}
+        for recommendation in recommendations:
+            group_key = str(recommendation.get("groupKey") or f"recommendation-{recommendation.get('id')}")
+            groups.setdefault(group_key, []).append(recommendation)
+        serialized: list[dict[str, Any]] = []
+        priority_order = {"p0": 0, "p1": 1, "p2": 2}
+        level_order = {"high": 0, "medium": 1, "low": 2}
+        for group_key, rows in groups.items():
+            quality = dict(rows[0].get("quality") or {})
+            change_set_ids = self._unique_strings([row.get("changeSetId") for row in rows if row.get("changeSetId")])
+            suggested_types = self._unique_strings(
+                item for row in rows for item in list(row.get("suggestedChangeTypes") or [])
+            )
+            serialized.append(
+                {
+                    "groupKey": group_key,
+                    "title": f"{quality.get('groupLabel') or group_key} 反馈组",
+                    "recommendationIds": [str(row.get("id")) for row in rows],
+                    "mergedCount": len(rows),
+                    "impactLevel": quality.get("impactLevel") or "medium",
+                    "riskLevel": quality.get("riskLevel") or "medium",
+                    "suggestedChangeTypes": suggested_types,
+                    "reviewPriority": quality.get("reviewPriority") or "p2",
+                    "changeSetCreated": bool(change_set_ids),
+                    "changeSetId": change_set_ids[0] if change_set_ids else None,
+                    "rationale": quality.get("rationale") or "基于同组发布后反馈聚合生成。",
+                }
+            )
+        return sorted(
+            serialized,
+            key=lambda row: (
+                priority_order.get(str(row.get("reviewPriority") or "p2"), 9),
+                level_order.get(str(row.get("riskLevel") or "medium"), 9),
+                str(row.get("groupKey") or ""),
+            ),
+        )
+
+    def _recommendation_quality_rationale(
+        self,
+        *,
+        focus_areas: list[str],
+        negative_feedback: float,
+        task_count: int,
+        priority: str,
+    ) -> str:
+        parts = []
+        if focus_areas:
+            parts.append("聚焦 " + "、".join(focus_areas[:4]))
+        if negative_feedback >= 1:
+            parts.append(f"负向反馈 {int(negative_feedback)} 条")
+        if task_count:
+            parts.append(f"建议任务 {task_count} 个")
+        parts.append(f"原始优先级 {priority or 'medium'}")
+        return "；".join(parts) + "。"
+
     def _iteration_recommendation_message(
         self,
         *,
@@ -2092,6 +2458,244 @@ class PMFlowOSProjectionService:
             )
         return items
 
+    def _iteration_group_message(
+        self,
+        *,
+        recommendations: list[IterationRecommendation],
+        feedback_by_id: dict[int, FeedbackIngest],
+        quality_by_id: dict[int, dict[str, Any]],
+        note: Optional[str],
+    ) -> str:
+        first_quality = next(iter(quality_by_id.values()), {})
+        summaries = [self._recommendation_summary_text(item) for item in recommendations]
+        feedback_titles = [
+            feedback_by_id[int(item.feedback_ingest_id)].title or feedback_by_id[int(item.feedback_ingest_id)].content[:60]
+            for item in recommendations
+            if item.feedback_ingest_id is not None and int(item.feedback_ingest_id) in feedback_by_id
+        ]
+        lines = [
+            "Release/Growth 同组反馈已生成下一轮待确认变更草稿。",
+            f"建议组：{first_quality.get('groupLabel') or first_quality.get('groupKey') or 'feedback'}",
+            f"合并建议：{len(recommendations)} 条；风险：{first_quality.get('riskLevel') or 'medium'}；影响：{first_quality.get('impactLevel') or 'medium'}。",
+        ]
+        if feedback_titles:
+            lines.append("反馈来源：" + "；".join(feedback_titles[:5]))
+        lines.append("建议摘要：" + "；".join(summaries[:4]))
+        if note:
+            lines.append(f"人工备注：{note.strip()}")
+        lines.append("此消息只作为 Change Review 证据，不自动写入 Project Memory。")
+        return "\n".join(lines)
+
+    def _iteration_group_change_set_headline(
+        self,
+        *,
+        recommendations: list[IterationRecommendation],
+        quality_by_id: dict[int, dict[str, Any]],
+    ) -> str:
+        quality = next(iter(quality_by_id.values()), {})
+        label = quality.get("groupLabel") or quality.get("groupKey") or "feedback"
+        return f"下一轮反馈变更草稿组：{label}（{len(recommendations)} 条建议）"[:255]
+
+    def _iteration_group_change_set_summary(
+        self,
+        *,
+        recommendations: list[IterationRecommendation],
+        quality_by_id: dict[int, dict[str, Any]],
+    ) -> str:
+        quality = next(iter(quality_by_id.values()), {})
+        return (
+            f"基于 {len(recommendations)} 条同组 iteration recommendation 生成下一轮 ChangeSet draft；"
+            f"reviewPriority={quality.get('reviewPriority') or 'p2'}，risk={quality.get('riskLevel') or 'medium'}，"
+            "等待 Change Review accept/edit/defer。"
+        )
+
+    def _iteration_group_excerpt(
+        self,
+        *,
+        recommendations: list[IterationRecommendation],
+        feedback_by_id: dict[int, FeedbackIngest],
+    ) -> str:
+        excerpts: list[str] = []
+        for recommendation in recommendations:
+            if recommendation.feedback_ingest_id is not None and int(recommendation.feedback_ingest_id) in feedback_by_id:
+                excerpts.append(feedback_by_id[int(recommendation.feedback_ingest_id)].content.strip())
+            else:
+                excerpts.append(self._recommendation_summary_text(recommendation))
+        return "\n".join(excerpts)[:1200]
+
+    def _iteration_recommendation_group_change_items(
+        self,
+        *,
+        recommendations: list[IterationRecommendation],
+        feedback_by_id: dict[int, FeedbackIngest],
+        quality_by_id: dict[int, dict[str, Any]],
+        project_id: int,
+    ) -> list[PMFlowChangeItem]:
+        first_quality = next(iter(quality_by_id.values()), {})
+        group_key = str(first_quality.get("groupKey") or "general-feedback")
+        recommendation_ids = [str(item.id) for item in recommendations]
+        feedback_ids = [str(item.feedback_ingest_id) for item in recommendations if item.feedback_ingest_id is not None]
+        focus_areas = self._unique_strings(
+            area
+            for recommendation in recommendations
+            for area in list(dict(recommendation.proposed_charter_patch or {}).get("focus_areas") or dict(recommendation.recommendation_summary or {}).get("focus_areas") or [])
+        )
+        summaries = [self._recommendation_summary_text(item) for item in recommendations]
+        summary_text = "；".join(summaries[:4])
+        risk_level = "high" if any(str(quality.get("riskLevel")) == "high" for quality in quality_by_id.values()) else "medium"
+        impact_level = "high" if any(str(quality.get("impactLevel")) == "high" for quality in quality_by_id.values()) else "medium"
+        base_payload = {
+            "source": "iteration_recommendation_group",
+            "groupKey": group_key,
+            "recommendationIds": recommendation_ids,
+            "feedbackIngestIds": feedback_ids,
+            "quality": first_quality,
+        }
+        items: list[PMFlowChangeItem] = [
+            PMFlowChangeItem(
+                tenant_id=self.tenant_id,
+                change_set_id=0,
+                project_id=int(project_id),
+                group="new_object",
+                object_type="Requirement",
+                object_id=None,
+                title=f"下一轮迭代方向：{first_quality.get('groupLabel') or group_key}"[:255],
+                summary=summary_text or "基于发布后反馈建议组生成下一轮迭代方向。",
+                diff="合并同组发布后反馈，形成下一轮需求方向与成功标准草稿。",
+                payload={
+                    **base_payload,
+                    "title": f"下一轮迭代方向：{first_quality.get('groupLabel') or group_key}",
+                    "summary": summary_text,
+                    "stage": "planning",
+                    "status": "draft",
+                    "focusAreas": focus_areas,
+                    "acceptanceCriteria": [
+                        f"本组 {len(recommendations)} 条 recommendation 在 Change Review 中完成 accept/edit/defer。",
+                        "接受后再进入 Project Memory，不自动写入。",
+                    ],
+                },
+                edited_payload=None,
+                confidence=0.78,
+                risk_level=risk_level,
+                status="pending_review",
+            ),
+            PMFlowChangeItem(
+                tenant_id=self.tenant_id,
+                change_set_id=0,
+                project_id=int(project_id),
+                group="decision",
+                object_type="Decision",
+                object_id=None,
+                title="确认同组反馈进入下一轮迭代",
+                summary=f"确认 {group_key} 反馈组是否作为下一轮迭代输入；影响级别 {impact_level}，风险级别 {risk_level}。",
+                diff="批量生成只进入待确认中心，不直接提交项目记忆。",
+                payload={
+                    **base_payload,
+                    "title": "确认同组反馈进入下一轮迭代",
+                    "summary": summary_text,
+                    "stage": "growth",
+                    "owner": "PMFlow",
+                },
+                edited_payload=None,
+                confidence=0.74,
+                risk_level=risk_level,
+                status="pending_review",
+            ),
+        ]
+
+        seen_task_titles: set[str] = set()
+        for recommendation in recommendations:
+            for raw_task in list(recommendation.proposed_tasks or []):
+                if not isinstance(raw_task, dict):
+                    continue
+                title = str(raw_task.get("title") or raw_task.get("summary") or "").strip()
+                if not title or title in seen_task_titles:
+                    continue
+                seen_task_titles.add(title)
+                task_summary = str(raw_task.get("summary") or raw_task.get("description") or title).strip()
+                items.append(
+                    PMFlowChangeItem(
+                        tenant_id=self.tenant_id,
+                        change_set_id=0,
+                        project_id=int(project_id),
+                        group="new_object",
+                        object_type="Task",
+                        object_id=None,
+                        title=title[:255],
+                        summary=task_summary,
+                        diff="从同组发布后反馈批量生成下一轮任务草稿，待 Change Review 决定是否接受。",
+                        payload={
+                            **base_payload,
+                            "title": title,
+                            "description": task_summary,
+                            "summary": task_summary,
+                            "stage": "development",
+                            "status": "draft",
+                            "priority": raw_task.get("priority") or first_quality.get("reviewPriority") or "p2",
+                        },
+                        edited_payload=None,
+                        confidence=0.72,
+                        risk_level="medium",
+                        status="pending_review",
+                    )
+                )
+
+        if risk_level == "high":
+            items.append(
+                PMFlowChangeItem(
+                    tenant_id=self.tenant_id,
+                    change_set_id=0,
+                    project_id=int(project_id),
+                    group="risk",
+                    object_type="Risk",
+                    object_id=None,
+                    title="同组发布后反馈高风险",
+                    summary=f"{group_key} 反馈组存在高风险或高影响建议，需先完成人工审查。",
+                    diff="风险项只进入 Change Review，不直接写入项目记忆。",
+                    payload={
+                        **base_payload,
+                        "title": "同组发布后反馈高风险",
+                        "summary": summary_text,
+                        "stage": "planning",
+                        "level": "high",
+                        "mitigation": "在 Change Review 中确认分组范围、owner 和处理顺序后再提交。",
+                    },
+                    edited_payload=None,
+                    confidence=0.76,
+                    risk_level="high",
+                    status="pending_review",
+                )
+            )
+        return items
+
+    async def _mark_recommendations_for_group_change_set(
+        self,
+        *,
+        recommendations: list[IterationRecommendation],
+        change_set_id: int,
+        group_key: str,
+        current_user_id: Optional[int],
+        reused: bool,
+    ) -> None:
+        stamped_at = now_cn_naive().isoformat()
+        for recommendation in recommendations:
+            summary_payload = dict(recommendation.recommendation_summary or {})
+            summary_payload.update(
+                {
+                    "changeSetId": str(change_set_id),
+                    "changeSetCreated": True,
+                    "changeSetCreatedAt": summary_payload.get("changeSetCreatedAt") or stamped_at,
+                    "changeSetCreatedBy": summary_payload.get("changeSetCreatedBy")
+                    or (None if current_user_id is None else str(current_user_id)),
+                    "groupKey": group_key,
+                    "groupChangeSetId": str(change_set_id),
+                    "groupDraftReused": bool(reused),
+                }
+            )
+            recommendation.recommendation_summary = summary_payload
+            recommendation.status = "changeset_drafted"
+            flag_modified(recommendation, "recommendation_summary")
+
     def _serialize_revenue_event(self, event: RevenueEvent) -> dict[str, Any]:
         return {
             "id": str(event.id),
@@ -2128,6 +2732,45 @@ class PMFlowOSProjectionService:
             actions.append("继续沉淀 feedback / revenue / compliance 信号，为下一轮 iteration recommendation 做准备。")
         return actions
 
+    def _normalize_group_key(self, value: Any) -> str:
+        text = str(value or "").strip().lower()
+        chars: list[str] = []
+        previous_dash = False
+        for char in text:
+            if char.isascii() and char.isalnum():
+                chars.append(char)
+                previous_dash = False
+            elif not previous_dash:
+                chars.append("-")
+                previous_dash = True
+        normalized = "".join(chars).strip("-")
+        return normalized[:64] or "general-feedback"
+
+    def _safe_number(self, value: Any) -> float:
+        if value is None or value == "":
+            return 0.0
+        try:
+            return float(value)
+        except (TypeError, ValueError):
+            return 0.0
+
+    def _unique_strings(self, values: Any) -> list[str]:
+        result: list[str] = []
+        for value in values or []:
+            text = str(value or "").strip()
+            if text and text not in result:
+                result.append(text)
+        return result
+
+    def _parse_recommendation_ids(self, raw_ids: list[str]) -> set[int]:
+        parsed: set[int] = set()
+        for raw in raw_ids or []:
+            text = str(raw or "").strip()
+            if not text.isdigit():
+                raise ValueError("iteration_recommendation_not_found")
+            parsed.add(int(text))
+        return parsed
+
     def _optional_int(self, value: Optional[Any], *, error_code: str) -> Optional[int]:
         if value is None or str(value).strip() == "":
             return None

backend/tests/test_pmflow_os_projection_api.py

@@ -1112,6 +1112,83 @@ async def test_iteration_recommendation_creates_reusable_changeset_draft(db_env)
     assert tasks == []
 
 
+@pytest.mark.asyncio
+async def test_iteration_recommendation_group_batches_quality_and_reuses_changeset_draft(db_env) -> None:
+    session = db_env
+    user, project = await _seed_user_project(session)
+    app = _build_app(session, SimpleNamespace(id=user.id, tenant_id=1, role="MEMBER"))
+
+    async with AsyncClient(transport=ASGITransport(app=app), base_url="http://test") as client:
+        await client.post(
+            f"/pmflow-os/projects/{project.id}/release-feedback-ingests",
+            json={
+                "source_type": "manual",
+                "title": "回答质量反馈 A",
+                "content": "bad answer 多次不准，用户要求下一轮补齐失败样本。",
+                "metrics": {"negative_feedback_count": 4},
+            },
+        )
+        await client.post(
+            f"/pmflow-os/projects/{project.id}/release-feedback-ingests",
+            json={
+                "source_type": "manual",
+                "title": "回答质量反馈 B",
+                "content": "出现幻觉和不准回答，需要改进 prompt quality。",
+                "metrics": {"negative_feedback_count": 2},
+            },
+        )
+        summary_response = await client.get(f"/pmflow-os/projects/{project.id}/execution-summary")
+        release_signals = summary_response.json()["data"]["releaseSignals"]
+        group = release_signals["recommendationGroups"][0]
+        draft_response = await client.post(
+            f"/pmflow-os/projects/{project.id}/iteration-recommendation-groups/{group['groupKey']}/change-set-drafts",
+            json={"recommendation_ids": group["recommendationIds"], "note": "批量进入下一轮待确认"},
+        )
+        reuse_response = await client.post(
+            f"/pmflow-os/projects/{project.id}/iteration-recommendation-groups/{group['groupKey']}/change-set-drafts",
+            json={"recommendation_ids": group["recommendationIds"]},
+        )
+
+    assert summary_response.status_code == 200
+    recommendations = release_signals["recommendations"]
+    assert group["groupKey"] == "quality"
+    assert group["mergedCount"] == 2
+    assert group["impactLevel"] == "high"
+    assert group["riskLevel"] == "high"
+    assert group["reviewPriority"] == "p0"
+    assert set(group["suggestedChangeTypes"]).issuperset({"Requirement", "Decision", "Task", "Risk"})
+    assert {item["mergedCount"] for item in recommendations} == {2}
+    assert sum(1 for item in recommendations if item["duplicateOf"]) == 1
+    assert all(item["groupKey"] == "quality" for item in recommendations)
+
+    assert draft_response.status_code == 200
+    payload = draft_response.json()["data"]
+    detail = payload["changeSetDetail"]
+    assert detail["changeSet"]["status"] == "open"
+    assert detail["evidence"][0]["sourceType"] == "iteration_recommendation_group"
+    assert detail["evidence"][0]["metadata"]["groupKey"] == "quality"
+    assert set(detail["evidence"][0]["metadata"]["recommendationIds"]) == set(group["recommendationIds"])
+    object_types = {item["objectType"] for item in detail["items"]}
+    assert object_types.issubset({"Requirement", "Task", "Risk", "Decision"})
+    assert {"Requirement", "Decision", "Task", "Risk"}.issubset(object_types)
+    assert all(item["status"] == "pending_review" for item in detail["items"])
+    updated_groups = payload["executionSummary"]["releaseSignals"]["recommendationGroups"]
+    assert updated_groups[0]["changeSetCreated"] is True
+    assert updated_groups[0]["changeSetId"] == detail["changeSet"]["id"]
+    assert all(
+        item["changeSetId"] == detail["changeSet"]["id"]
+        for item in payload["executionSummary"]["releaseSignals"]["recommendations"]
+    )
+
+    assert reuse_response.status_code == 200
+    assert reuse_response.json()["data"]["changeSetDetail"]["changeSet"]["id"] == detail["changeSet"]["id"]
+    assert "iteration_recommendation_group_changeset_reused" in reuse_response.json()["data"]["warnings"]
+    change_sets = (await session.execute(select(PMFlowChangeSet).where(PMFlowChangeSet.project_id == project.id))).scalars().all()
+    tasks = (await session.execute(select(Task).where(Task.project_id == project.id))).scalars().all()
+    assert len(change_sets) == 1
+    assert tasks == []
+
+
 @pytest.mark.asyncio
 async def test_execution_progress_writeback_rejects_repo_outside_project(db_env) -> None:
     session = db_env

docs/PMFLOW_PRODUCT_DELIVERY_STANDARD.md

@@ -715,16 +715,13 @@ commit gate 判定：
 
 当前最建议下一步执行：
 
-**P7：Iteration Quality / Reviewer / Personal Queue 三线并行。**
+**P7：Iteration Quality / Reviewer / Personal Queue 收口。**
 
 原因：
 
-- A3 已完成真实只读投影，Phase B 已完成真实 Project Launch 创建链。
+- P6 后主流程已经形成真实闭环：Project Launch 创建项目、Change Review 审查并 commit 项目真相、Execution Bridge 交接与执行回写、Release Validation 记录发布前验证、Release Feedback 摄取发布后信号，并能把 `IterationRecommendation` 送回 Change Review。
-- Phase C1 已把项目内消息、结构化候选、审查动作和证据链落到真实表。
+- 当前缺口不再是 D.2 execution handoff，而是下一轮建议的质量治理、审查解释和个人队列承接：同类 recommendation 需要分组/去重/评估风险，Reviewer 需要知道“为什么生成这一组待确认变更”，Space/Personal Queue 需要把下一步审查动作排进个人工作面。
-- C2 已实现真实 ChangeSet commit，Phase D/D.2/D.3 已把 Execution Bridge 从真实读侧推进到执行交接和执行结果回写。
+- P7 不新增战略页面，不重开路由；只沿现有 `ReleaseGrowthPage / ChangeReviewPage / PersonalSpace` 主链补质量、审查和队列能力。
-- Release Validation P1、Release Feedback P1、Iteration Loop P1 已把发布前验证、发布后反馈和 recommendation -> 下一轮 ChangeSet draft 串回 Change Review。
-- Space P5 已只读展示下一轮迭代草稿关注，Account P6 已把 iteration review 事件映射到待确认提醒偏好。
-- 当前缺口不是继续扩页面或重做执行桥，而是让“下一轮待确认变更”更可审、更可分配、更容易被个人队列接住。
 
 ### 7.1 P7 并行建议
 
@@ -734,9 +731,8 @@ P7 不新增战略页面，不新增顶层路由，不改移动底部导航；
 
 目标：
 
-- 提升 recommendation -> ChangeSet draft 的质量：分组、去重、风险级别、证据摘要和 nextActions 更清晰。
+- 当前 Main Flow 下一步进入 `iteration-loop-p2-recommendation-quality-batch`：在 `execution-summary.releaseSignals` 中补 recommendation quality fields，支持同组 recommendation 批量生成一个可复用 ChangeSet draft，并在 Change Review 解释来源组、反馈关系、风险和审查优先级。
-- 让 ReleaseGrowthPage / ChangeReviewPage 能区分“反馈建议草稿、任务草稿、风险草稿、决策草稿”，但不自动 commit。
+- 后续 Reviewer packet 可继续围绕待确认项排序、review checklist、accept/edit/defer 质量说明推进，但不得新增战略页或自动写 Project Memory。
-- 保持 item 类型只落 `Requirement / Task / Risk / Decision`；不写 Project Memory、不创建正式 Task、不触发发布。
 
 建议写范围：
 
@@ -1029,7 +1025,7 @@ Shell / Integration owner 负责：
 未完成或风险：
 
 - 本阶段只接入 feedback ingest 与 releaseSignals 读模型；RevenueEvent / ComplianceCheckRun 已可被后续扩展继续纳入更完整发布后运营读模型。
-- iteration recommendation 目前作为建议底座展示，不自动生成 ChangeSet；下一轮如果要从 recommendation 进入待确认中心，需要单独开 Main Flow packet。
+- iteration recommendation 已能生成下一轮 ChangeSet draft；P7 继续推进 recommendation quality batch，使同类建议可分组、去重、评估风险并批量进入待确认中心。
 - 全量前端 typecheck/build 需在包含 `frontend/src/lib/*` helper 的集成环境或由 Shell / Integration owner 修正该 ignored 文件边界后重新验证。
 
 ### 2026-04-26 / iteration-loop-p1-recommendation-to-changeset-draft / implemented
@@ -1070,6 +1066,46 @@ Shell / Integration owner 负责：
 - 本阶段只把 release feedback recommendation 转入 Change Review 待确认中心；不执行部署、不触发外部支付/推送、不自动写 Project Memory、不自动创建正式任务。
 - Space 仍只读消费 `releaseSignals` 并跳转主流程；Account Settings 只提供偏好 helper，不参与 PMFlow OS 写入。
 
+### 2026-04-26 / iteration-loop-p2-recommendation-quality-batch / implemented
+
+完成内容：
+
+- `execution-summary.releaseSignals.recommendations` 新增 recommendation quality 字段：`groupKey / duplicateOf / mergedCount / impactLevel / riskLevel / suggestedChangeTypes / reviewPriority / quality`，并新增 `recommendationGroups` 聚合同组建议。
+- 新增 `POST /pmflow-os/projects/{project_id}/iteration-recommendation-groups/{group_key}/change-set-drafts`，支持按同组 recommendation 批量生成一个 open ChangeSet draft；重复调用会复用既有 draft 并返回 `iteration_recommendation_group_changeset_reused`。
+- 后端批量 draft 仍只生成 pending_review `Requirement / Task / Risk / Decision` ChangeItem 与 `PMFlowEvidenceLink(source_type='iteration_recommendation_group')`，不自动 commit、不写 Project Memory、不创建正式 Task。
+- 已生成 draft 的 recommendation 会在 `recommendation_summary` 追踪 `groupKey / groupChangeSetId / changeSetId / changeSetCreated`，使 releaseSignals 与 Change Review 能解释同组反馈来源。
+- `ReleaseGrowthPage` 新增 recommendation group 卡片，展示合并数量、影响、风险、审查优先级、建议变更类型与生成理由，并调用真实批量 API；单条 recommendation 动作保留。
+- `ChangeReviewPage` 在打开来自 release feedback iteration group 的 ChangeSet 时展示来源组、recommendation 数、feedback 数、风险/影响/审查优先级与 rationale，帮助 reviewer 判断为什么生成这一组待确认变更。
+
+涉及文件：
+
+- `backend/app/api/pmflow_os_api/__init__.py`
+- `backend/app/schemas/pmflow_os.py`
+- `backend/app/services/pmflow_os_projection_service.py`
+- `backend/tests/test_pmflow_os_projection_api.py`
+- `frontend/src/api/pmflowOs.ts`
+- `frontend/src/features/pmflow-os/store.tsx`
+- `frontend/src/features/pmflow-os/store.api.test.tsx`
+- `frontend/src/features/pmflow-os/pages/ReleaseGrowthPage.tsx`
+- `frontend/src/features/pmflow-os/pages/ChangeReviewPage.tsx`
+- `docs/PMFLOW_PRODUCT_DELIVERY_STANDARD.md`
+- `docs/architecture/CODE_REFERENCE_INDEX.md`
+- `docs/architecture/ARCHITECTURE_INVENTORY.md`
+- `CLAUDE.md`
+
+验证：
+
+- `PYTHONPATH=backend /opt/homebrew/bin/python3.11 -m py_compile backend/app/api/pmflow_os_api/__init__.py backend/app/services/pmflow_os_projection_service.py backend/app/schemas/pmflow_os.py` 通过。
+- `PYTHONPATH=backend /opt/homebrew/bin/python3.11 -m pytest backend/tests/test_pmflow_os_projection_api.py -q` 通过：`20 passed`；仍有既有 `backend/app/schemas/recipe.py` Pydantic V2 class-based config deprecation warning。
+- `cd frontend && npm run test -- src/features/pmflow-os` 通过：`2 files / 10 tests passed`。
+- `cd frontend && npm run typecheck` 未通过：当前独立 worktree 缺少被 `.gitignore` 排除的 `frontend/src/lib/*` 本地 helper 文件，导致全仓 `@/lib/api`、`@/lib/utils`、`@/lib/datetime`、`@/lib/display`、`@/lib/rbac` 等 alias 解析失败；这些文件不在本 Packet write scope，未越界补入。
+- `cd frontend && npm run build` 同样停在上述 `frontend/src/lib/*` 缺失的 TypeScript 解析错误，未进入 Vite build。
+
+边界：
+
+- 本阶段只提升下一轮建议质量与批量待确认能力；不执行部署、不接外部支付、不接推送、不自动写 Project Memory、不创建正式 Task。
+- Space / Account / router / bottom nav 未改；Personal Queue 承接可在后续 P7 packet 中单独推进。
+
 ### 2026-04-26 / big-module-integration-p4 / done
 
 完成内容：

docs/architecture/ARCHITECTURE_INVENTORY.md

@@ -197,9 +197,9 @@
 - `frontend/src/features/personal-space/PersonalSpacePage.tsx` 是空间首页视图层，`frontend/src/api/space.ts` 是空间工作面只读聚合 API：从 `projects` summary、PMFlow OS `truth-state` 与 `execution-summary` 生成继续推进、待确认、发布关注、发布后反馈关注、下一轮迭代草稿关注、执行中 / AI 正在推进、暂存想法、草案箱、最近项目摘要；Space P3 的 `releaseAttention` 优先消费 `truth-state.releases` 中由 release validation run 投影出的 release 状态与检查项，把“发布验证阻塞 / 发布验证待确认”显示到 `/space`，没有发布验证投影时才回落到 `execution-summary.releaseGate` 的发布候选或发布门禁阻塞；Space P4/P5 的 `feedbackAttention` 只读消费 `execution-summary.releaseSignals`，把发布后 feedback、revenue、ROI、compliance、iteration recommendation 与已生成的下一轮 ChangeSet draft 关注项显示到 `/space`。发布验证、发布后信号或下一轮 draft 缺失时显示明确空态，不虚构验证、反馈、收入、ROI、合规、建议或 ChangeSet draft 结果。项目主动作进入 `/pmflow-os?projectId=<id>`，待确认项进入 `/change-review?projectId=<id>&changeSetId=<id>`，发布关注和发布后反馈关注项进入 `/release-growth?projectId=<id>`，下一轮 draft 关注项进入 `/change-review?projectId=<id>&changeSetId=<id>`，执行中项进入 `/execution-bridge?projectId=<id>`；暂存想法后端只读聚合尚未接入时只显示 empty/warning，不使用 mock 数据冒充真实内容；移动端为空间首页独立 App 布局，不直接压缩桌面卡片，并保留最近项目区块；`frontend/src/pages/PersonalSpacePage.tsx` 是薄入口。该域只表达个人空间和主流程导航，不做团队空间强切换、企业组织管理、Project Memory 写入、ChangeSet 状态写入、ProgressEvent 写入、ReleaseValidationRun 写入、FeedbackIngest 写入或发布执行。
 - `frontend/src/features/project-launch/types.ts`、`mockData.ts`、`useProjectLaunchState.ts` 是新项目启动台状态层；默认真实模式通过 `frontend/src/api/pmflowLaunch.ts` 生成草案并确认创建真实项目，API 失败进入明确失败态，不静默 fallback 成 mock 成功；`mockData.ts` 仅在显式 `?demo=1` 或 `?dataSource=demo` 下作为演示状态机使用。`components/LaunchShell.tsx` 负责桌面/移动分流，桌面端复用 `frontend/src/features/personal-space/components/SpaceChrome.tsx` 的空间左侧栏和顶部条，`LaunchCanvases.tsx` 和 `MobileLaunchLayout.tsx` 承接高保真图中的意图输入、AI 工作、人类确认、草案修改、生成失败、创建成功过渡与移动端 bottom sheet 交互。
 - `frontend/src/features/account-settings/AccountSettingsPanel.tsx` 是 `/me` 设置页实现：桌面端为个人资料摘要 + 分区设置面板，移动端为分组列表 + `MobileBottomSheet` 编辑；当前覆盖个人资料、默认语言、AI 回答深度、证据展示偏好、待确认、执行回写、发布门禁和失败提醒通知偏好，并在通知偏好区展示页面内提醒消费说明。`frontend/src/api/accountSettings.ts` 是前端合同层；有 token 时优先调用 `backend/app/api/account_settings_api/`，失败时回落浏览器本机暂存；P2 新增 `shouldShowAccountNotification` 与 `getAccountNotificationConsumption`，将四类通知偏好映射为本地 UI 的待确认、Execution Bridge、Release/Growth 和失败提醒显示/静默状态；P3 新增 `getReleaseGateNotificationPreference(settings, event)`，把 Release/Growth 的 `release_gate_* / release_validation_*` 事件映射到 `releaseGate` 偏好，把 `failed_event` 映射到 `failureAlert` 偏好；P5/P6 的 `getPostReleaseNotificationPreference(settings, event)` 把发布后 `feedback_received / revenue_signal / compliance_blocked` 映射到 `releaseGate` 偏好，把 `iteration_recommendation / next_change_review` 映射到 `pendingConfirmation` 偏好和 `change_review` surface，并返回 `spaceReadable / evidencePolicy / reminderPolicy` 供 Main Flow 与 Space 消费。该合同只控制页面内提醒表达，不代表邮件、桌面推送或系统通知中心已接入；偏好关闭也不得隐藏业务证据、recommendation 或 ChangeSet。`backend/app/schemas/account_settings.py` 定义账号偏好合同，`backend/app/services/account_settings_service.py` 当前复用既有 `integrations` 表按 `tenant_id + user_id` 保存 `account_settings_preferences` 配置包，`backend/app/api/router.py` 已挂载该 API；未改用户表、OAuth/JWT/auth core 或 PMFlow OS truth store。
-- `frontend/src/features/pmflow-os/store.tsx` 当前是 PMFlow OS 共享 truth store：`?projectId=<id>` 时通过 React Query 调用 `frontend/src/api/pmflowOs.ts` 读取真实 `GET /pmflow-os/projects/{project_id}/truth-state` 与 `GET /pmflow-os/projects/{project_id}/execution-summary`，真实模式下项目内 message、`proposed_items` extraction、ChangeItem `accept/edit/defer`、ChangeSet commit、Execution Handoff create/consume、execution progress writeback、release validation run create、release feedback ingest 与 iteration recommendation -> ChangeSet draft 均调用 `/pmflow-os` 后端并 refetch；只有显式 `?demo=1` 或 `?dataSource=demo` 时才启用本地 mock 演示状态机。真实读取/动作失败会进入错误/空态，不静默 fallback 成 mock 成功。
+- `frontend/src/features/pmflow-os/store.tsx` 当前是 PMFlow OS 共享 truth store：`?projectId=<id>` 时通过 React Query 调用 `frontend/src/api/pmflowOs.ts` 读取真实 `GET /pmflow-os/projects/{project_id}/truth-state` 与 `GET /pmflow-os/projects/{project_id}/execution-summary`，真实模式下项目内 message、`proposed_items` extraction、ChangeItem `accept/edit/defer`、ChangeSet commit、Execution Handoff create/consume、execution progress writeback、release validation run create、release feedback ingest、单条 iteration recommendation -> ChangeSet draft 与同组 iteration recommendation 批量 -> ChangeSet draft 均调用 `/pmflow-os` 后端并 refetch；只有显式 `?demo=1` 或 `?dataSource=demo` 时才启用本地 mock 演示状态机。真实读取/动作失败会进入错误/空态，不静默 fallback 成 mock 成功。
-- `frontend/src/features/pmflow-os/store.tsx` 仍保留演示用项目内 AI 结构化闭环状态机：`command_idle / ai_extracting / extraction_ready / review_open / writeback_success`。`frontend/src/features/pmflow-os/pages/StructuredExtractionPage.tsx` 是本轮结构化提取页，`ChangeReviewPage.tsx` 是决策与待确认中心，`MemoryWritebackResultPage.tsx` 是项目记忆写入结果页，`WorkGraphPage.tsx` 是项目工作台关系总览与对象投影视图，`ExecutionBridgePage.tsx` 是执行桥上下文包、交接按钮、运行端领取确认、PR/测试与回写状态页；`ReleaseGrowthPage.tsx` 真实模式读取 `execution-summary.releaseGate`、`latestValidationRun` 与 `releaseSignals`，从 branch、commit、PR、test、failed evidence 展示 `可发布 / 阻塞 / 需要人工确认 / 缺少证据` 判断、阻塞原因、下一步动作、发布验证记录和发布后 feedback/revenue/recommendation 信号，缺少发布后信号时显示 empty/warning，不执行发布命令、不展示 mock 发布结果；页面可把真实 recommendation 生成下一轮待确认 ChangeSet draft，生成后仍进入 Change Review accept/edit/defer，不自动 commit、不写 Project Memory、不创建正式任务；同时读取 Account notification preference helper，用 `releaseGate` 偏好控制发布门禁强调提醒，用 `failureAlert` 偏好控制失败强调提醒，偏好关闭时只静默提醒表达强度，不隐藏 release evidence、阻塞原因、审计状态、validation run 或 releaseSignals；静态发布/增长原型只保留在显式 demo 模式。A3 已接入后端 truth-state，C1 已接入真实 message/extraction/review 状态机，C2 已接入真实 ChangeSet commit，Phase D 已接入真实 execution-summary/dev-sync，D.2 已接入 `POST /execution-handoffs` 创建 PromptPackage + handoff_created progress event，P1 已接入 `POST /execution-handoffs/{prompt_package_id}/consume` 创建 handoff_consumed progress event 并把 execution status 投影为 coding，Release Growth P1 已接入 execution evidence gate，Release Validation P1 已接入 `POST /release-validation-runs` 创建 `ReleaseValidationRun` 快照，Release Feedback P1 已接入 `POST /release-feedback-ingests` 写入 `FeedbackIngest` 并生成 `IterationRecommendation`，Iteration Loop P1 已接入 `POST /iteration-recommendations/{recommendation_id}/change-set-drafts` 把 recommendation 回流到 Change Review；当前仍不调用真实 AI 提取，不把 mock 当真实失败 fallback。`ProjectMemoryPage.tsx` 在真实模式下展示后端 `truth_nodes + semantic_facts` 的 `projectMemory` 投影，accepted/edited ChangeItem 只有 commit 成功后才进入 Project Memory。`frontend/src/features/pmflow-os/components/ProjectStructuredLoop.tsx` 提供 `StatusLabel / EvidenceSnippet / ObjectChangeCard / ProjectConversationPanel / StructuredExtractionPanel / ProjectAICopilotPanel` 等闭环组件，并复用契约化 `InlineNotice` 表达“未确认不得写入”的局部提示。
+- `frontend/src/features/pmflow-os/store.tsx` 仍保留演示用项目内 AI 结构化闭环状态机：`command_idle / ai_extracting / extraction_ready / review_open / writeback_success`。`frontend/src/features/pmflow-os/pages/StructuredExtractionPage.tsx` 是本轮结构化提取页，`ChangeReviewPage.tsx` 是决策与待确认中心，`MemoryWritebackResultPage.tsx` 是项目记忆写入结果页，`WorkGraphPage.tsx` 是项目工作台关系总览与对象投影视图，`ExecutionBridgePage.tsx` 是执行桥上下文包、交接按钮、运行端领取确认、PR/测试与回写状态页；`ReleaseGrowthPage.tsx` 真实模式读取 `execution-summary.releaseGate`、`latestValidationRun` 与 `releaseSignals`，从 branch、commit、PR、test、failed evidence 展示 `可发布 / 阻塞 / 需要人工确认 / 缺少证据` 判断、阻塞原因、下一步动作、发布验证记录和发布后 feedback/revenue/recommendation 信号，缺少发布后信号时显示 empty/warning，不执行发布命令、不展示 mock 发布结果；页面可展示 recommendation quality grouping，把真实 recommendation 单条或同组批量生成下一轮待确认 ChangeSet draft，生成后仍进入 Change Review accept/edit/defer，不自动 commit、不写 Project Memory、不创建正式任务；`ChangeReviewPage.tsx` 会展示 `iteration_recommendation_group` evidence 的来源组、feedback 数、recommendation 数、风险、影响和审查优先级；同时读取 Account notification preference helper，用 `releaseGate` 偏好控制发布门禁强调提醒，用 `failureAlert` 偏好控制失败强调提醒，偏好关闭时只静默提醒表达强度，不隐藏 release evidence、阻塞原因、审计状态、validation run 或 releaseSignals；静态发布/增长原型只保留在显式 demo 模式。A3 已接入后端 truth-state，C1 已接入真实 message/extraction/review 状态机，C2 已接入真实 ChangeSet commit，Phase D 已接入真实 execution-summary/dev-sync，D.2 已接入 `POST /execution-handoffs` 创建 PromptPackage + handoff_created progress event，P1 已接入 `POST /execution-handoffs/{prompt_package_id}/consume` 创建 handoff_consumed progress event 并把 execution status 投影为 coding，Release Growth P1 已接入 execution evidence gate，Release Validation P1 已接入 `POST /release-validation-runs` 创建 `ReleaseValidationRun` 快照，Release Feedback P1 已接入 `POST /release-feedback-ingests` 写入 `FeedbackIngest` 并生成 `IterationRecommendation`，Iteration Loop P1 已接入 `POST /iteration-recommendations/{recommendation_id}/change-set-drafts` 把 recommendation 回流到 Change Review，Iteration Loop P2 已接入 `POST /iteration-recommendation-groups/{group_key}/change-set-drafts` 并在 `releaseSignals.recommendationGroups` 暴露分组/去重/风险/影响合同；当前仍不调用真实 AI 提取，不把 mock 当真实失败 fallback。`ProjectMemoryPage.tsx` 在真实模式下展示后端 `truth_nodes + semantic_facts` 的 `projectMemory` 投影，accepted/edited ChangeItem 只有 commit 成功后才进入 Project Memory。`frontend/src/features/pmflow-os/components/ProjectStructuredLoop.tsx` 提供 `StatusLabel / EvidenceSnippet / ObjectChangeCard / ProjectConversationPanel / StructuredExtractionPanel / ProjectAICopilotPanel` 等闭环组件，并复用契约化 `InlineNotice` 表达“未确认不得写入”的局部提示。
-- PMFlow OS 后端入口位于 `backend/app/api/pmflow_os_api/`，projection service 位于 `backend/app/services/pmflow_os_projection_service.py`，Change Review / commit 服务位于 `backend/app/services/pmflow_change_review_service.py`，Pydantic 合同位于 `backend/app/schemas/pmflow_os.py`；当前聚合 `truth_nodes / tasks / semantic_facts / ai_project_control / dev_sync`，`Thread / Message` 来自 `pmflow_threads / pmflow_thread_messages`，`ChangeSet / ChangeItem / Approval / EvidenceLink` 来自 `backend/app/models/pmflow_change_review.py` 对应真实表。C1 已实现 message、`proposed_items` extraction、ChangeSet detail、ChangeItem `accept/edit/defer`；C2 backend/frontend 已实现 `POST /pmflow-os/change-sets/{change_set_id}/commit` 与真实模式前端接入，默认 atomic，将 `Requirement / Decision / Risk` 写入 `truth_nodes`，将 `Task` 写入 `truth_nodes(node_type='task') + tasks`，高风险项强制 approval gate；D.2 已实现 `POST /pmflow-os/projects/{project_id}/execution-handoffs`，校验 task/repo project 边界，生成/复用 `PromptPackage(package_kind='execution_handoff')` 并写 `ProgressEvent(event_type='handoff_created')` 起点；P1 已实现 `POST /pmflow-os/projects/{project_id}/execution-handoffs/{prompt_package_id}/consume`，校验 execution handoff package 后更新 `PromptPackage.package_payload.latest_handoff.handoff_state=consumed` 并写 `ProgressEvent(event_type='handoff_consumed', derived_task_status='coding')`；Release Growth P1 在 `ExecutionSummaryProjection.releaseGate` 中从 `ProgressEvent` 推导发布门禁状态，失败事件 fail-closed 阻塞发布，无证据返回 empty/warning；Release Validation P1 新增 `POST /pmflow-os/projects/{project_id}/release-validation-runs`，复用现有 `ReleaseValidationRun` 模型固化当前 `releaseGate` 快照、人工决策、证据、阻塞原因和下一步动作，并在 `releaseGate.latestValidationRun` 返回最新记录；Release Feedback P1 新增 `POST /pmflow-os/projects/{project_id}/release-feedback-ingests`，复用现有 `FeedbackIngest / IterationRecommendation`，在 `ExecutionSummaryProjection.releaseSignals` 中返回真实发布后 feedback、revenue event 和 iteration recommendation 读模型；Iteration Loop P1 新增 `POST /pmflow-os/projects/{project_id}/iteration-recommendations/{recommendation_id}/change-set-drafts`，把真实 recommendation 转为 open ChangeSet、pending_review Requirement/Task/Risk/Decision items 与 `iteration_recommendation` evidence，重复调用复用既有 draft 并在 `releaseSignals` 标记 `changeSetCreated/changeSetId`；这些步骤都不直接执行本机命令、真实发布命令、外部支付、真实推送、自动 Project Memory commit 或正式任务创建；`ModelDecision / MetricSignal` 在 C2 V1 仍 fail-closed。
+- PMFlow OS 后端入口位于 `backend/app/api/pmflow_os_api/`，projection service 位于 `backend/app/services/pmflow_os_projection_service.py`，Change Review / commit 服务位于 `backend/app/services/pmflow_change_review_service.py`，Pydantic 合同位于 `backend/app/schemas/pmflow_os.py`；当前聚合 `truth_nodes / tasks / semantic_facts / ai_project_control / dev_sync`，`Thread / Message` 来自 `pmflow_threads / pmflow_thread_messages`，`ChangeSet / ChangeItem / Approval / EvidenceLink` 来自 `backend/app/models/pmflow_change_review.py` 对应真实表。C1 已实现 message、`proposed_items` extraction、ChangeSet detail、ChangeItem `accept/edit/defer`；C2 backend/frontend 已实现 `POST /pmflow-os/change-sets/{change_set_id}/commit` 与真实模式前端接入，默认 atomic，将 `Requirement / Decision / Risk` 写入 `truth_nodes`，将 `Task` 写入 `truth_nodes(node_type='task') + tasks`，高风险项强制 approval gate；D.2 已实现 `POST /pmflow-os/projects/{project_id}/execution-handoffs`，校验 task/repo project 边界，生成/复用 `PromptPackage(package_kind='execution_handoff')` 并写 `ProgressEvent(event_type='handoff_created')` 起点；P1 已实现 `POST /pmflow-os/projects/{project_id}/execution-handoffs/{prompt_package_id}/consume`，校验 execution handoff package 后更新 `PromptPackage.package_payload.latest_handoff.handoff_state=consumed` 并写 `ProgressEvent(event_type='handoff_consumed', derived_task_status='coding')`；Release Growth P1 在 `ExecutionSummaryProjection.releaseGate` 中从 `ProgressEvent` 推导发布门禁状态，失败事件 fail-closed 阻塞发布，无证据返回 empty/warning；Release Validation P1 新增 `POST /pmflow-os/projects/{project_id}/release-validation-runs`，复用现有 `ReleaseValidationRun` 模型固化当前 `releaseGate` 快照、人工决策、证据、阻塞原因和下一步动作，并在 `releaseGate.latestValidationRun` 返回最新记录；Release Feedback P1 新增 `POST /pmflow-os/projects/{project_id}/release-feedback-ingests`，复用现有 `FeedbackIngest / IterationRecommendation`，在 `ExecutionSummaryProjection.releaseSignals` 中返回真实发布后 feedback、revenue event 和 iteration recommendation 读模型；Iteration Loop P1 新增 `POST /pmflow-os/projects/{project_id}/iteration-recommendations/{recommendation_id}/change-set-drafts`，把真实 recommendation 转为 open ChangeSet、pending_review Requirement/Task/Risk/Decision items 与 `iteration_recommendation` evidence，重复调用复用既有 draft 并在 `releaseSignals` 标记 `changeSetCreated/changeSetId`；Iteration Loop P2 新增 `POST /pmflow-os/projects/{project_id}/iteration-recommendation-groups/{group_key}/change-set-drafts`，把同组 recommendation 批量转成一个 open ChangeSet draft 与 `iteration_recommendation_group` evidence，并在 `releaseSignals.recommendations / recommendationGroups` 投影 `groupKey / duplicateOf / mergedCount / impactLevel / riskLevel / suggestedChangeTypes / reviewPriority`；这些步骤都不直接执行本机命令、真实发布命令、外部支付、真实推送、自动 Project Memory commit 或正式任务创建；`ModelDecision / MetricSignal` 在 C2 V1 仍 fail-closed。
 - PMFlow Launch 后端 Phase B 创建链位于 `backend/app/api/pmflow_launch_api/`、`backend/app/services/pmflow_launch_service.py`、`backend/app/schemas/pmflow_launch.py` 与 `backend/app/models/pmflow_launch.py`；`POST /pmflow-launch/drafts` 只写 `pmflow_launch_drafts`，`POST /pmflow-launch/drafts/{draft_id}/confirm` 在同一事务内写 `projects / user_projects / pmflow_threads / pmflow_thread_messages / truth_nodes / tasks / semantic_facts seed / operation_logs`，并把 `launch_draft_id / thread_id / thread_message_id` 写入正式对象 metadata。
 - `frontend/src/features/pmflow-os/components/ChangeReviewGroup.tsx`、`WorkbenchPanel.tsx`、`ExecutionBridgeCards.tsx`、`PMFlowOsFrame.tsx`、`AccountControls.tsx`、`FeedbackPrimitives.tsx` 是高保真原型新增的行为组件层：分别承载变更 diff 审查块、工作台对象卡组、执行桥上下文/日志卡片、全局上下文条与 AI 副驾、账号菜单/退出登录确认交互，以及 `PageBanner / InlineNotice / HumanDecisionBar / ConfirmDialog / MobileBottomSheet` 这组提示、提醒、确认与移动端 bottom sheet 基础组件，避免再把这些业务片段内联塞回 `pages/`。
 - `frontend/vite.config.ts` 现补了固定 vendor chunk 分组（`framework / state-query / ui-radix / icons / charts / markdown`）与业务 chunk 分组（`ai-steward / sstc / governance`），目标是让登录页与基础导航不再背 AI Steward、SSTC、治理页和图表代码。

docs/architecture/CODE_REFERENCE_INDEX.md

@@ -89,7 +89,7 @@
 - SSTC 负责人推荐接口：参考 `POST /projects/{project_id}/tasks/{task_id}/recommend-assignee`
 - SSTC 任务更新接口：参考 `PATCH /projects/{project_id}/tasks/{task_id}`
 - SSTC 发布创建接口：参考 `POST /projects/{project_id}/releases`
-- PMFlow OS 真实读路径、C1 Change Review API、C2 backend commit、Execution Handoff、Release Evidence Gate、Release Validation Run、Release Feedback Signal 与 Iteration Loop draft：参考 `backend/app/api/pmflow_os_api/`、`backend/app/services/pmflow_os_projection_service.py`、`backend/app/services/pmflow_change_review_service.py`、`backend/app/schemas/pmflow_os.py`；只读提供 `GET /pmflow-os/projects/{project_id}/truth-state` 与 `GET /pmflow-os/projects/{project_id}/execution-summary`，复用 `truth_nodes / tasks / semantic_facts / ai_project_control / dev_sync`，并读取 `pmflow_threads / pmflow_thread_messages / pmflow_change_sets / pmflow_change_items / pmflow_approvals / pmflow_evidence_links`；C1 写 API 包括 thread message、`proposed_items` extraction、ChangeSet detail、ChangeItem `accept/edit/defer`；C2 backend 已实现 `POST /pmflow-os/change-sets/{change_set_id}/commit`，默认 `mode=atomic`，将 `Requirement / Decision / Risk` 写入 `truth_nodes`，将 `Task` 写入 `truth_nodes(node_type='task') + tasks`，高风险项强制 approval gate；D.2 已实现 `POST /pmflow-os/projects/{project_id}/execution-handoffs`，对已确认任务生成/复用 `PromptPackage` 并创建 `ProgressEvent(event_type='handoff_created')` 起点；P1 已实现 `POST /pmflow-os/projects/{project_id}/execution-handoffs/{prompt_package_id}/consume`，记录运行端领取、更新 `PromptPackage.package_payload.latest_handoff.handoff_state=consumed` 并创建 `ProgressEvent(event_type='handoff_consumed', derived_task_status='coding')`；Release Growth P1 已在 `execution-summary.releaseGate` 中从 branch / commit / PR / test_passed / checks_passed / failed / blocked 等 `ProgressEvent` 推导 `ready / blocked / needs_confirmation / empty`；Release Validation P1 新增 `POST /pmflow-os/projects/{project_id}/release-validation-runs`，从当前 `releaseGate` 固化 `ReleaseValidationRun` 快照并在 `releaseGate.latestValidationRun` 返回最新记录，无 evidence 时返回 `release_validation_evidence_empty`；Release Feedback P1 新增 `POST /pmflow-os/projects/{project_id}/release-feedback-ingests`，复用 `FeedbackIngest / IterationRecommendation` 写入真实发布后反馈，并在 `execution-summary.releaseSignals` 投影真实 `feedback / revenueEvents / recommendations` 与 release gate / validation run 关系；Iteration Loop P1 新增 `POST /pmflow-os/projects/{project_id}/iteration-recommendations/{recommendation_id}/change-set-drafts`，把真实 `IterationRecommendation` 转为 open ChangeSet、pending_review Requirement/Task/Risk/Decision items 与 iteration evidence，重复调用复用既有 draft；这些步骤均不执行发布命令、外部支付、真实推送或自动 Project Memory commit
+- PMFlow OS 真实读路径、C1 Change Review API、C2 backend commit、Execution Handoff、Release Evidence Gate、Release Validation Run、Release Feedback Signal 与 Iteration Loop draft：参考 `backend/app/api/pmflow_os_api/`、`backend/app/services/pmflow_os_projection_service.py`、`backend/app/services/pmflow_change_review_service.py`、`backend/app/schemas/pmflow_os.py`；只读提供 `GET /pmflow-os/projects/{project_id}/truth-state` 与 `GET /pmflow-os/projects/{project_id}/execution-summary`，复用 `truth_nodes / tasks / semantic_facts / ai_project_control / dev_sync`，并读取 `pmflow_threads / pmflow_thread_messages / pmflow_change_sets / pmflow_change_items / pmflow_approvals / pmflow_evidence_links`；C1 写 API 包括 thread message、`proposed_items` extraction、ChangeSet detail、ChangeItem `accept/edit/defer`；C2 backend 已实现 `POST /pmflow-os/change-sets/{change_set_id}/commit`，默认 `mode=atomic`，将 `Requirement / Decision / Risk` 写入 `truth_nodes`，将 `Task` 写入 `truth_nodes(node_type='task') + tasks`，高风险项强制 approval gate；D.2 已实现 `POST /pmflow-os/projects/{project_id}/execution-handoffs`，对已确认任务生成/复用 `PromptPackage` 并创建 `ProgressEvent(event_type='handoff_created')` 起点；P1 已实现 `POST /pmflow-os/projects/{project_id}/execution-handoffs/{prompt_package_id}/consume`，记录运行端领取、更新 `PromptPackage.package_payload.latest_handoff.handoff_state=consumed` 并创建 `ProgressEvent(event_type='handoff_consumed', derived_task_status='coding')`；Release Growth P1 已在 `execution-summary.releaseGate` 中从 branch / commit / PR / test_passed / checks_passed / failed / blocked 等 `ProgressEvent` 推导 `ready / blocked / needs_confirmation / empty`；Release Validation P1 新增 `POST /pmflow-os/projects/{project_id}/release-validation-runs`，从当前 `releaseGate` 固化 `ReleaseValidationRun` 快照并在 `releaseGate.latestValidationRun` 返回最新记录，无 evidence 时返回 `release_validation_evidence_empty`；Release Feedback P1 新增 `POST /pmflow-os/projects/{project_id}/release-feedback-ingests`，复用 `FeedbackIngest / IterationRecommendation` 写入真实发布后反馈，并在 `execution-summary.releaseSignals` 投影真实 `feedback / revenueEvents / recommendations` 与 release gate / validation run 关系；Iteration Loop P1 新增 `POST /pmflow-os/projects/{project_id}/iteration-recommendations/{recommendation_id}/change-set-drafts`，把真实 `IterationRecommendation` 转为 open ChangeSet、pending_review Requirement/Task/Risk/Decision items 与 iteration evidence，重复调用复用既有 draft；Iteration Loop P2 新增 `POST /pmflow-os/projects/{project_id}/iteration-recommendation-groups/{group_key}/change-set-drafts`，并在 `execution-summary.releaseSignals.recommendations` 补 `groupKey / duplicateOf / mergedCount / impactLevel / riskLevel / suggestedChangeTypes / reviewPriority`，在 `recommendationGroups` 聚合同组建议，支持同组 recommendation 批量生成一个可复用 ChangeSet draft 与 `iteration_recommendation_group` evidence；这些步骤均不执行发布命令、外部支付、真实推送、自动 Project Memory commit 或正式 Task 创建
 - Project Launch 真实创建链：参考 `backend/app/api/pmflow_launch_api/`、`backend/app/services/pmflow_launch_service.py`、`backend/app/schemas/pmflow_launch.py` 与模型 `backend/app/models/pmflow_launch.py`；接口为 `POST /pmflow-launch/drafts` 与 `POST /pmflow-launch/drafts/{draft_id}/confirm`，draft 只写 `pmflow_launch_drafts`，confirm 在同一事务内创建 `projects / user_projects / pmflow_threads / pmflow_thread_messages / truth_nodes / tasks / semantic_facts seed / operation_logs`
 - PMFlow Change Review 模型与 migration：参考 `backend/app/models/pmflow_change_review.py` 与 `backend/alembic/versions/20260426_add_pmflow_change_review_tables.py`；只包含 `pmflow_change_sets / pmflow_change_items / pmflow_approvals / pmflow_evidence_links`，本阶段不得新增或依赖 `/commit`
 
@@ -98,7 +98,7 @@
 - PMFlow 大模块并行公共壳：参考 `docs/PMFLOW_PRODUCT_DELIVERY_STANDARD.md` 的 `4.4.1-4.4.3` 与 `frontend/src/router/index.tsx`、`frontend/src/main.tsx`、`frontend/src/features/pmflow-os/components/MobileBottomNav.tsx`、`frontend/src/features/pmflow-os/components/PMFlowOsFrame.tsx`；`/space` 归 Space lane，`/pmflow-os / /command-center / /change-review / /execution-bridge / /release-growth` 归 Main Flow lane，`/me` 归 Account Settings lane，`/settings` 当前仅为 reserved route，新增或改动必须由 Shell / Integration owner 串行处理
 - PMFlow P7 当前推荐任务：参考 `docs/PMFLOW_PRODUCT_DELIVERY_STANDARD.md` 第 7 节；P6 后真实闭环已到 `feedback -> recommendation -> ChangeSet draft -> Change Review`，下一步推荐并行推进 `main-flow-p7-iteration-quality`、`main-flow-p7-reviewer-decision-desk`、`space-p7-personal-queue`，不新增战略页面、不改 router / bottom nav、不绕过 Change Review 人类确认边界
 - PMFlow OS 产品合同真源：先读 `docs/architecture/PMFLOW_OS_PRODUCT_CONTRACT.md`，再实现 `pmflow-os` 真实后端接入；该文档锁定 `ProjectTruthState / Thread / ChangeSet / ChangeItem / Evidence / Approval / Project Memory / ExecutionContext` 与 `truth_nodes / tasks / semantic_facts / ai_project_control / dev_sync` 的复用边界和后续 API 草案
-- 新增前端 API 调用：参考 `frontend/src/api/sstc.ts`；Space 工作面只读聚合参考 `frontend/src/api/space.ts`，只消费项目摘要与 PMFlow OS `truth-state / execution-summary`，不写 Project Memory、ChangeSet 状态、ProgressEvent 或 ReleaseValidationRun；Space P3 的 `releaseAttention` 优先从 `truth-state.releases` 的 release validation 投影生成“发布验证阻塞 / 发布验证待确认”，没有发布验证投影时才回落到 `PMFlowExecutionSummary.releaseGate` 的“发布候选 / 发布门禁阻塞”，所有动作只跳转 `/release-growth?projectId=<id>`；PMFlow OS 真实接入参考 `frontend/src/api/pmflowOs.ts` 的 `getPMFlowTruthState(projectId)`、`getPMFlowExecutionSummary(projectId)`、`createPMFlowExecutionHandoff`、`consumePMFlowExecutionHandoff`、`createPMFlowExecutionProgressEvent`、`createPMFlowReleaseValidationRun`、`createPMFlowReleaseFeedbackIngest`、`createPMFlowIterationRecommendationChangeSetDraft`、`createPMFlowThreadMessage`、`createPMFlowExtraction`、`getPMFlowChangeSet`、`acceptPMFlowChangeItem`、`editPMFlowChangeItem`、`deferPMFlowChangeItem`、`commitPMFlowChangeSet`；`PMFlowExecutionSummary.releaseGate` 是 ReleaseGrowth 真实发布门禁前端合同，`releaseGate.latestValidationRun` 是最新发布验证记录前端合同，`execution-summary.releaseSignals` 是 Release/Growth 发布后 feedback/revenue/recommendation 信号与下一轮 ChangeSet draft 状态前端合同；Project Launch 创建链参考 `frontend/src/api/pmflowLaunch.ts` 的 `createPMFlowLaunchDraft` 与 `confirmPMFlowLaunchDraft`
+- 新增前端 API 调用：参考 `frontend/src/api/sstc.ts`；Space 工作面只读聚合参考 `frontend/src/api/space.ts`，只消费项目摘要与 PMFlow OS `truth-state / execution-summary`，不写 Project Memory、ChangeSet 状态、ProgressEvent 或 ReleaseValidationRun；Space P3 的 `releaseAttention` 优先从 `truth-state.releases` 的 release validation 投影生成“发布验证阻塞 / 发布验证待确认”，没有验证投影时才回落到 `PMFlowExecutionSummary.releaseGate` 的“发布候选 / 发布门禁阻塞”，所有动作只跳转 `/release-growth?projectId=<id>`；PMFlow OS 真实接入参考 `frontend/src/api/pmflowOs.ts` 的 `getPMFlowTruthState(projectId)`、`getPMFlowExecutionSummary(projectId)`、`createPMFlowExecutionHandoff`、`consumePMFlowExecutionHandoff`、`createPMFlowExecutionProgressEvent`、`createPMFlowReleaseValidationRun`、`createPMFlowReleaseFeedbackIngest`、`createPMFlowIterationRecommendationChangeSetDraft`、`createPMFlowIterationRecommendationGroupChangeSetDraft`、`createPMFlowThreadMessage`、`createPMFlowExtraction`、`getPMFlowChangeSet`、`acceptPMFlowChangeItem`、`editPMFlowChangeItem`、`deferPMFlowChangeItem`、`commitPMFlowChangeSet`；`PMFlowExecutionSummary.releaseGate` 是 ReleaseGrowth 真实发布门禁前端合同，`releaseGate.latestValidationRun` 是最新发布验证记录前端合同，`execution-summary.releaseSignals` 是 Release/Growth 发布后 feedback/revenue/recommendation 信号、recommendation quality grouping 与下一轮 ChangeSet draft 状态前端合同；Project Launch 创建链参考 `frontend/src/api/pmflowLaunch.ts` 的 `createPMFlowLaunchDraft` 与 `confirmPMFlowLaunchDraft`
 - AI Steward 页面入口：参考 `frontend/src/pages/ai-steward/`
 - AI Steward 问答浮窗与上下文显性化组件：参考 `frontend/src/components/ai-steward/FloatingQAChat.tsx`、`frontend/src/components/ai-steward/QAConversationStatusBar.tsx`、`frontend/src/components/ai-steward/QAResponseMetaCard.tsx`
 - AI Steward QA contract 前端类型（含 `analysis_packet`）：参考 `frontend/src/api/aiSteward.ts`
@@ -108,8 +108,8 @@
 - 个人空间首页 / 空间层入口前端产品域：参考 `frontend/src/features/personal-space/`、`frontend/src/api/space.ts` 与薄入口 `frontend/src/pages/PersonalSpacePage.tsx`；本地预览路由入口为 `/space`，覆盖继续推进、待我确认、发布关注、发布后反馈关注、下一轮迭代草稿关注、执行中 / AI 正在推进、暂存想法、草案箱与最近项目，点击项目主动作进入 `/pmflow-os?projectId=<id>`，待确认项进入 `/change-review?projectId=<id>&changeSetId=<id>`，执行中项进入 `/execution-bridge?projectId=<id>`，发布关注和发布后反馈关注项进入 `/release-growth?projectId=<id>`，已生成 ChangeSet draft 的下一轮建议进入 `/change-review?projectId=<id>&changeSetId=<id>`；Space 只读聚合项目摘要、PMFlow OS `truth-state.releases / execution-summary.releaseGate / execution-summary.releaseSignals`，发布验证、发布后信号或 iteration draft 缺失时显示明确空态，不虚构验证、反馈、收入、ROI、合规、iteration recommendation 或 ChangeSet draft，不直接写 Project Memory、ChangeSet 状态、execution event、ReleaseValidationRun、FeedbackIngest、RevenueEvent 或 ComplianceCheckRun；暂存想法后端只读聚合尚未接入时必须显示 empty/warning，不得用 mock 数据冒充真实内容；空间层共享左侧栏、顶部条与空间用量组件在 `frontend/src/features/personal-space/components/SpaceChrome.tsx`，`/project-launch` 桌面首次新建壳也复用该组件
 - 新项目启动台 / 首次进入页前端产品域：参考 `frontend/src/features/project-launch/` 与薄入口 `frontend/src/pages/ProjectLaunchPage.tsx`；真实模式默认通过 `frontend/src/api/pmflowLaunch.ts` 调用后端 draft/confirm API，只有显式 `?demo=1` 或 `?dataSource=demo` 才启用 `mockData.ts` 本地演示状态机；本地预览路由入口为 `/project-launch`
 - Account Settings / `/me` 前端产品域与 P1/P2/P3/P5/P6 偏好合同：参考 `frontend/src/features/account-settings/AccountSettingsPanel.tsx`、`frontend/src/features/pmflow-os/pages/AccountPage.tsx`、`frontend/src/api/accountSettings.ts`、`backend/app/api/account_settings_api/`、`backend/app/schemas/account_settings.py` 与 `backend/app/services/account_settings_service.py`；当前覆盖个人资料、默认语言、AI 回答深度、证据展示偏好、待确认、执行回写、发布门禁和失败提醒通知偏好，移动端使用分组列表与 bottom sheet，账户退出仍复用 `AccountControls.tsx`，不改 OAuth/JWT/auth core；P1 后端合同提供 `GET /account-settings/me` 与 `PUT /account-settings/me`，已由 `backend/app/api/router.py` 挂入核心路由组，按当前 `tenant_id + user_id` 保存账号偏好，前端有 token 时优先账号同步，后端不可用时明确回落 `仅本机暂存 / 保存失败`；P2 在 `frontend/src/api/accountSettings.ts` 提供 `shouldShowAccountNotification` 与 `getAccountNotificationConsumption`，把通知偏好映射为页面内提醒消费模型；P3 新增 `getReleaseGateNotificationPreference(settings, event)` 供 Release/Growth 后续消费：`release_gate_*` 与 `release_validation_*` 使用 `releaseGate` 偏好，`failed_event` 使用 `failureAlert` 偏好；P5/P6 的 `getPostReleaseNotificationPreference(settings, event)` 供发布后反馈与下一轮待确认消费：`feedback_received / revenue_signal / compliance_blocked` 使用 `releaseGate` 偏好，`iteration_recommendation / next_change_review` 使用 `pendingConfirmation` 偏好并落到 `change_review` surface，返回 `spaceReadable / showInAppReminder / evidencePolicy / reminderPolicy`；ReleaseGrowthPage 当前消费 `releaseGate` 与 `failureAlert` 两个偏好，只静默强调型提醒，不隐藏 release evidence、阻塞原因、审计状态、validation run、releaseSignals 或下一轮 ChangeSet draft；这些 helper 只控制本地 UI 提醒消费合同，不代表真实邮件、桌面推送或系统通知中心已接入
-- PMFlow OS 共享 truth store 与主链状态动作：参考 `frontend/src/features/pmflow-os/store.tsx`；当前支持 `?projectId=<id>` 走 React Query 读取真实 `truth-state` 与 `execution-summary`，真实模式下 thread message、`proposed_items` extraction、ChangeItem `accept/edit/defer`、ChangeSet commit、Execution Handoff create/consume、execution progress writeback、release validation run create、release feedback ingest 与 iteration recommendation -> ChangeSet draft 均调后端并 refetch，不本地 promote 入 Project Memory、不伪造 execution event、发布验证记录、发布后信号或下一轮 draft；`?demo=1` 才启用本地 mock 演示状态机和本地动作；项目内 AI 结构化闭环状态机为 `command_idle / ai_extracting / extraction_ready / review_open / writeback_success`
+- PMFlow OS 共享 truth store 与主链状态动作：参考 `frontend/src/features/pmflow-os/store.tsx`；当前支持 `?projectId=<id>` 走 React Query 读取真实 `truth-state` 与 `execution-summary`，真实模式下 thread message、`proposed_items` extraction、ChangeItem `accept/edit/defer`、ChangeSet commit、Execution Handoff create/consume、execution progress writeback、release validation run create、release feedback ingest、单条 iteration recommendation -> ChangeSet draft 与同组 iteration recommendation 批量 -> ChangeSet draft 均调后端并 refetch，不本地 promote 入 Project Memory、不伪造 execution event、发布验证记录、发布后信号或下一轮 draft；`?demo=1` 才启用本地 mock 演示状态机和本地动作；项目内 AI 结构化闭环状态机为 `command_idle / ai_extracting / extraction_ready / review_open / writeback_success`
-- PMFlow OS 产品页入口：参考 `frontend/src/features/pmflow-os/pages/CommandCenterPage.tsx`、`StructuredExtractionPage.tsx`、`ChangeReviewPage.tsx`、`MemoryWritebackResultPage.tsx`、`WorkGraphPage.tsx`、`ExecutionBridgePage.tsx`、`ReleaseGrowthPage.tsx`、`TeamLayerPage.tsx`、`ProjectMemoryPage.tsx`、`AccountPage.tsx`；`ReleaseGrowthPage.tsx` 真实模式展示 `execution-summary.releaseGate`、execution evidence、阻塞原因、下一步动作、`latestValidationRun` 和 `releaseSignals`，缺证据或缺发布后信号时显示 empty/warning，不展示 mock 发布结果；页面可调用真实 API 创建 release validation run、摄取 release feedback，并从真实 recommendation 生成下一轮待确认 ChangeSet draft，生成后仍进入 Change Review accept/edit/defer，不自动 commit、不写 Project Memory、不执行部署命令、不接外部支付；同时消费 Account notification preference helper，`releaseGate` 关闭时静默发布门禁强调提醒，`failureAlert` 关闭时静默失败强调提醒，所有真实证据和审计状态仍可见；静态高保真发布/增长数据仅保留在显式 demo 模式
+- PMFlow OS 产品页入口：参考 `frontend/src/features/pmflow-os/pages/CommandCenterPage.tsx`、`StructuredExtractionPage.tsx`、`ChangeReviewPage.tsx`、`MemoryWritebackResultPage.tsx`、`WorkGraphPage.tsx`、`ExecutionBridgePage.tsx`、`ReleaseGrowthPage.tsx`、`TeamLayerPage.tsx`、`ProjectMemoryPage.tsx`、`AccountPage.tsx`；`ReleaseGrowthPage.tsx` 真实模式展示 `execution-summary.releaseGate`、execution evidence、阻塞原因、下一步动作、`latestValidationRun` 和 `releaseSignals`，缺证据或缺发布后信号时显示 empty/warning，不展示 mock 发布结果；页面可调用真实 API 创建 release validation run、摄取 release feedback，从真实 recommendation 生成下一轮待确认 ChangeSet draft，并按 `recommendationGroups` 批量生成同组下一轮 ChangeSet draft；`ChangeReviewPage.tsx` 会解释来自 `iteration_recommendation_group` evidence 的组、反馈、风险、影响和审查优先级；生成后仍进入 Change Review accept/edit/defer，不自动 commit、不写 Project Memory、不执行部署命令、不接外部支付；同时消费 Account notification preference helper，`releaseGate` 关闭时静默发布门禁强调提醒，`failureAlert` 关闭时静默失败强调提醒，所有真实证据和审计状态仍可见；静态高保真发布/增长数据仅保留在显式 demo 模式
 - PMFlow OS 行为组件：参考 `frontend/src/features/pmflow-os/components/ProjectStructuredLoop.tsx`、`ProjectStageRail.tsx`、`ThreadComposer.tsx`、`ExtractionRail.tsx`、`ChangeGroupPanel.tsx`、`ObjectGraphSection.tsx`、`ContextPacketPanel.tsx`、`ExecutionTimeline.tsx`、`ReleaseGatePanel.tsx`、`TeamFeedPanel.tsx`、`ChangeReviewGroup.tsx`、`WorkbenchPanel.tsx`、`ExecutionBridgeCards.tsx`、`PMFlowOsFrame.tsx`、`AccountControls.tsx`、`FeedbackPrimitives.tsx`、`MobileBottomNav.tsx`；其中 `ProjectStructuredLoop.tsx` 承接 `StatusLabel / EvidenceSnippet / ObjectChangeCard / ProjectConversationPanel / StructuredExtractionPanel / ProjectAICopilotPanel`，`PMFlowOsFrame.tsx` 承接项目内面包屑和项目切换器，`AccountControls.tsx` 承接桌面头像菜单、退出登录确认与移动端 bottom sheet，`FeedbackPrimitives.tsx` 承接 `PageBanner / InlineNotice / HumanDecisionBar / ConfirmDialog / MobileBottomSheet`，`MobileBottomNav.tsx` 承接正式应用壳移动端固定底部导航
 - PMFlow OS 原型状态回归测试：参考 `frontend/src/features/pmflow-os/store.test.ts`；真实 API 动作回归参考 `frontend/src/features/pmflow-os/store.api.test.tsx`
 - Project Launch 真实/演示分支回归测试：参考 `frontend/src/features/project-launch/useProjectLaunchState.test.ts`

frontend/src/api/pmflowOs.ts

@@ -164,6 +164,7 @@ export interface PMFlowReleaseSignals {
   feedback: PMFlowReleaseFeedbackIngest[];
   revenueEvents: PMFlowReleaseRevenueEvent[];
   recommendations: PMFlowIterationRecommendation[];
+  recommendationGroups?: PMFlowIterationRecommendationGroup[];
   nextActions: string[];
   warnings: string[];
 }
@@ -208,10 +209,32 @@ export interface PMFlowIterationRecommendation {
   proposedTasks: Array<Record<string, unknown>>;
   changeSetCreated: boolean;
   changeSetId?: string | null;
+  groupKey?: string | null;
+  duplicateOf?: string | null;
+  mergedCount?: number;
+  impactLevel?: string | null;
+  riskLevel?: string | null;
+  suggestedChangeTypes?: string[];
+  reviewPriority?: string | null;
+  quality?: Record<string, unknown>;
   createdAt?: string | null;
   updatedAt?: string | null;
 }
 
+export interface PMFlowIterationRecommendationGroup {
+  groupKey: string;
+  title?: string | null;
+  recommendationIds: string[];
+  mergedCount: number;
+  impactLevel: string;
+  riskLevel: string;
+  suggestedChangeTypes: string[];
+  reviewPriority: string;
+  changeSetCreated: boolean;
+  changeSetId?: string | null;
+  rationale?: string | null;
+}
+
 export interface PMFlowReleaseFeedbackIngestRequest {
   source_type?: string;
   title?: string | null;
@@ -233,6 +256,11 @@ export interface PMFlowIterationRecommendationChangeSetDraftRequest {
   note?: string | null;
 }
 
+export interface PMFlowIterationRecommendationGroupChangeSetDraftRequest {
+  recommendation_ids?: string[];
+  note?: string | null;
+}
+
 export interface PMFlowIterationRecommendationChangeSetDraftResult {
   changeSetDetail: PMFlowChangeSetDetail;
   executionSummary: PMFlowExecutionSummary;
@@ -445,6 +473,17 @@ export async function createPMFlowIterationRecommendationChangeSetDraft(
   );
 }
 
+export async function createPMFlowIterationRecommendationGroupChangeSetDraft(
+  projectId: string,
+  groupKey: string,
+  payload: PMFlowIterationRecommendationGroupChangeSetDraftRequest = {}
+): Promise<PMFlowIterationRecommendationChangeSetDraftResult> {
+  return post<PMFlowIterationRecommendationChangeSetDraftResult>(
+    `/pmflow-os/projects/${projectId}/iteration-recommendation-groups/${encodeURIComponent(groupKey)}/change-set-drafts`,
+    payload
+  );
+}
+
 export async function createPMFlowThreadMessage(
   projectId: string,
   threadId: string,

frontend/src/features/pmflow-os/pages/ChangeReviewPage.tsx

@@ -9,12 +9,43 @@ import { usePMFlowOsStore } from "@/features/pmflow-os/store";
 
 const { Filter, ListFilter } = reviewToolbarIcons;
 
+type IterationEvidenceMeta = {
+  sourceType?: string;
+  changeSetId?: string;
+  metadata?: {
+    groupKey?: string;
+    recommendationIds?: string[];
+    feedbackIngestIds?: string[];
+    quality?: {
+      impactLevel?: string;
+      riskLevel?: string;
+      reviewPriority?: string;
+      rationale?: string;
+    };
+  };
+};
+
+function getIterationEvidence(state: ReturnType<typeof usePMFlowOsStore>["state"], changeSetId?: string | null) {
+  if (!changeSetId) return null;
+  return (
+    state.evidence.find((entry) => {
+      const raw = entry as typeof entry & IterationEvidenceMeta;
+      return (
+        raw.changeSetId === changeSetId &&
+        (raw.sourceType === "iteration_recommendation_group" || raw.sourceType === "iteration_recommendation")
+      );
+    }) || null
+  );
+}
+
 export default function ChangeReviewPage() {
   const navigate = useNavigate();
   const { state, dataSourceMode, acceptChange, editChange, deferChange, promoteAcceptedToGraph } = usePMFlowOsStore();
   const activeChangeSet = state.changeSets.find((entry) => entry.id === state.activeChangeSetId) || state.changeSets[0] || null;
   const items = state.changeItems.filter((item) => item.changeSetId === activeChangeSet?.id);
   const selectedItems = items.filter((item) => item.status === "pending_review").slice(0, 2);
+  const iterationEvidence = getIterationEvidence(state, activeChangeSet?.id);
+  const iterationMeta = (iterationEvidence as (typeof iterationEvidence & IterationEvidenceMeta) | null)?.metadata;
 
   const writeback = () => {
     if (!activeChangeSet) return;
@@ -47,6 +78,23 @@ export default function ChangeReviewPage() {
           </div>
 
           <ReviewExtractionHero activeChangeSet={activeChangeSet} />
+          {iterationEvidence ? (
+            <div className="rounded-xl border border-indigo-100 bg-indigo-50 p-4 text-sm leading-6 text-indigo-800">
+              <div className="font-semibold">
+                {iterationMeta?.groupKey ? "来自发布后反馈建议组" : "来自发布后反馈建议"}
+              </div>
+              <div className="mt-1">
+                {iterationMeta?.groupKey ? `组 ${iterationMeta.groupKey}；` : ""}
+                recommendation {iterationMeta?.recommendationIds?.length || 1} 条；feedback{" "}
+                {iterationMeta?.feedbackIngestIds?.length || 0} 条；风险{" "}
+                {iterationMeta?.quality?.riskLevel || "未评估"}；影响 {iterationMeta?.quality?.impactLevel || "未评估"}；审查{" "}
+                {iterationMeta?.quality?.reviewPriority || "p2"}。
+              </div>
+              <div className="mt-1 text-xs text-indigo-600">
+                {iterationMeta?.quality?.rationale || "这些变更只进入 Change Review，接受前不会写入 Project Memory 或创建正式 Task。"}
+              </div>
+            </div>
+          ) : null}
           <ChangeReviewTable
             items={items}
             evidence={state.evidence}

frontend/src/features/pmflow-os/pages/ReleaseGrowthPage.tsx

@@ -23,6 +23,7 @@ import { useState, type ButtonHTMLAttributes, type ReactNode } from "react";
 import { getAccountSettings, shouldShowAccountNotification } from "@/api/accountSettings";
 import type {
   PMFlowIterationRecommendation,
+  PMFlowIterationRecommendationGroup,
   PMFlowReleaseFeedbackIngest,
   PMFlowReleaseGateEvidence,
   PMFlowReleaseRevenueEvent,
@@ -394,6 +395,98 @@ function recommendationText(item: PMFlowIterationRecommendation) {
   return typeof text === "string" && text.trim() ? text : "已基于真实反馈生成下一轮建议，等待进入后续 iteration recommendation 主链。";
 }
 
+function qualityLabel(value?: string | null) {
+  if (value === "high") return "高";
+  if (value === "medium") return "中";
+  if (value === "low") return "低";
+  return value || "未评估";
+}
+
+function priorityLabel(value?: string | null) {
+  if (value === "p0") return "P0";
+  if (value === "p1") return "P1";
+  if (value === "p2") return "P2";
+  return value || "P2";
+}
+
+function qualityTone(value?: string | null): StatusTone {
+  if (value === "high" || value === "p0") return "danger";
+  if (value === "medium" || value === "p1") return "warning";
+  if (value === "low" || value === "p2") return "success";
+  return "neutral";
+}
+
+function RecommendationQualityMeta({ item }: { item: PMFlowIterationRecommendation }) {
+  return (
+    <div className="mt-3 flex flex-wrap gap-2 text-xs">
+      {item.groupKey ? <StatusChip label={`组 ${item.groupKey}`} tone="info" /> : null}
+      <StatusChip label={`影响 ${qualityLabel(item.impactLevel)}`} tone={qualityTone(item.impactLevel)} />
+      <StatusChip label={`风险 ${qualityLabel(item.riskLevel)}`} tone={qualityTone(item.riskLevel)} />
+      <StatusChip label={`审查 ${priorityLabel(item.reviewPriority)}`} tone={qualityTone(item.reviewPriority)} />
+      {item.mergedCount && item.mergedCount > 1 ? <StatusChip label={`合并 ${item.mergedCount} 条`} tone="info" /> : null}
+      {item.duplicateOf ? <StatusChip label={`重复于 #${item.duplicateOf}`} tone="neutral" /> : null}
+      {(item.suggestedChangeTypes || []).slice(0, 4).map((type) => (
+        <span key={type} className="rounded-md border border-blue-200 bg-white px-2 py-1 text-blue-600">
+          {type}
+        </span>
+      ))}
+    </div>
+  );
+}
+
+function RecommendationGroupRow({
+  group,
+  isLoading,
+  onCreate,
+}: {
+  group: PMFlowIterationRecommendationGroup;
+  isLoading: boolean;
+  onCreate: (group: PMFlowIterationRecommendationGroup) => void;
+}) {
+  return (
+    <div className="rounded-lg border border-indigo-100 bg-indigo-50 px-4 py-3 text-sm leading-6 text-indigo-800">
+      <div className="flex flex-col gap-3 md:flex-row md:items-start md:justify-between">
+        <div className="min-w-0">
+          <div className="font-semibold">{group.title || `反馈建议组 · ${group.groupKey}`}</div>
+          <div className="mt-1 text-indigo-700">
+            {group.rationale || "基于同组发布后反馈聚合生成；用于决定是否批量进入下一轮 Change Review。"}
+          </div>
+          <div className="mt-3 flex flex-wrap gap-2 text-xs">
+            <StatusChip label={`组 ${group.groupKey}`} tone="info" />
+            <StatusChip label={`合并 ${group.mergedCount} 条`} tone="info" />
+            <StatusChip label={`影响 ${qualityLabel(group.impactLevel)}`} tone={qualityTone(group.impactLevel)} />
+            <StatusChip label={`风险 ${qualityLabel(group.riskLevel)}`} tone={qualityTone(group.riskLevel)} />
+            <StatusChip label={`审查 ${priorityLabel(group.reviewPriority)}`} tone={qualityTone(group.reviewPriority)} />
+            {group.suggestedChangeTypes.slice(0, 4).map((type) => (
+              <span key={type} className="rounded-md border border-indigo-200 bg-white px-2 py-1 text-indigo-600">
+                {type}
+              </span>
+            ))}
+          </div>
+          {group.changeSetCreated ? (
+            <div className="mt-2 text-xs text-indigo-600">
+              已复用 ChangeSet draft {group.changeSetId ? `#${group.changeSetId}` : ""}，同组 recommendation 会追踪到同一个 draft。
+            </div>
+          ) : (
+            <div className="mt-2 text-xs text-indigo-600">批量生成只进入 Change Review，不自动 commit、不写 Project Memory、不创建正式 Task。</div>
+          )}
+        </div>
+        <Button
+          className={cn(
+            "min-h-9 shrink-0 whitespace-normal rounded-lg px-3 py-2 text-center text-xs leading-4 md:whitespace-nowrap",
+            group.changeSetCreated ? "border border-indigo-200 bg-white text-indigo-500" : "bg-indigo-600 text-white hover:bg-indigo-700"
+          )}
+          disabled={isLoading || group.changeSetCreated}
+          onClick={() => onCreate(group)}
+        >
+          {group.changeSetCreated ? <CheckCircle2 className="mr-1.5 h-3.5 w-3.5" /> : <ArrowRight className="mr-1.5 h-3.5 w-3.5" />}
+          {group.changeSetCreated ? "已生成组草稿" : "批量生成这一组待确认变更"}
+        </Button>
+      </div>
+    </div>
+  );
+}
+
 function checkTone(status?: string): StatusTone {
   if (status === "passed") return "success";
   if (status === "failed") return "danger";
@@ -494,6 +587,7 @@ function RealReleaseGrowthPage({ store }: { store: PMFlowOsStoreSnapshot }) {
     createReleaseValidationRun,
     createReleaseFeedbackIngest,
     createIterationRecommendationChangeSetDraft,
+    createIterationRecommendationGroupChangeSetDraft,
   } = store;
   const gate = executionSummary?.releaseGate;
   const releaseSignals = executionSummary?.releaseSignals;
@@ -542,6 +636,13 @@ function RealReleaseGrowthPage({ store }: { store: PMFlowOsStoreSnapshot }) {
       note: "ReleaseGrowthPage 从发布后反馈建议生成下一轮待确认 ChangeSet draft。",
     });
   };
+  const handleCreateRecommendationGroupChangeSet = (group: PMFlowIterationRecommendationGroup) => {
+    if (group.changeSetCreated) return;
+    createIterationRecommendationGroupChangeSetDraft(group.groupKey, {
+      recommendation_ids: group.recommendationIds,
+      note: `ReleaseGrowthPage 从 ${group.groupKey} 同组发布后反馈批量生成下一轮待确认 ChangeSet draft。`,
+    });
+  };
 
   return (
     <div className="min-h-full bg-[#f7faff] text-slate-950">
@@ -678,12 +779,16 @@ function RealReleaseGrowthPage({ store }: { store: PMFlowOsStoreSnapshot }) {
                 {(releaseSignals?.revenueEvents || []).map((item) => (
                   <RevenueSignalRow key={item.id} item={item} />
                 ))}
+                {(releaseSignals?.recommendationGroups || []).map((group) => (
+                  <RecommendationGroupRow key={group.groupKey} group={group} isLoading={isLoading} onCreate={handleCreateRecommendationGroupChangeSet} />
+                ))}
                 {(releaseSignals?.recommendations || []).map((item) => (
                   <div key={item.id} className="rounded-lg border border-blue-100 bg-blue-50 px-4 py-3 text-sm leading-6 text-blue-700">
                     <div className="flex flex-col gap-3 md:flex-row md:items-start md:justify-between">
                       <div>
                         <div className="font-semibold">下一轮建议 · {item.priority}</div>
                         <div className="mt-1">{recommendationText(item)}</div>
+                        <RecommendationQualityMeta item={item} />
                         {item.changeSetCreated ? (
                           <div className="mt-2 text-xs text-blue-600">
                             已生成 ChangeSet draft {item.changeSetId ? `#${item.changeSetId}` : ""}，请到 Change Review 继续 accept / edit / defer。

frontend/src/features/pmflow-os/store.api.test.tsx

@@ -10,6 +10,7 @@ import {
   consumePMFlowExecutionHandoff,
   createPMFlowExecutionHandoff,
   createPMFlowExecutionProgressEvent,
+  createPMFlowIterationRecommendationGroupChangeSetDraft,
   createPMFlowIterationRecommendationChangeSetDraft,
   createPMFlowReleaseValidationRun,
   createPMFlowReleaseFeedbackIngest,
@@ -48,6 +49,7 @@ vi.mock("@/api/pmflowOs", () => ({
   consumePMFlowExecutionHandoff: vi.fn(),
   createPMFlowExecutionHandoff: vi.fn(),
   createPMFlowExecutionProgressEvent: vi.fn(),
+  createPMFlowIterationRecommendationGroupChangeSetDraft: vi.fn(),
   createPMFlowIterationRecommendationChangeSetDraft: vi.fn(),
   createPMFlowReleaseValidationRun: vi.fn(),
   createPMFlowReleaseFeedbackIngest: vi.fn(),
@@ -62,6 +64,7 @@ const createThreadMessageMock = vi.mocked(createPMFlowThreadMessage);
 const createExecutionHandoffMock = vi.mocked(createPMFlowExecutionHandoff);
 const consumeExecutionHandoffMock = vi.mocked(consumePMFlowExecutionHandoff);
 const createExecutionProgressEventMock = vi.mocked(createPMFlowExecutionProgressEvent);
+const createIterationRecommendationGroupChangeSetDraftMock = vi.mocked(createPMFlowIterationRecommendationGroupChangeSetDraft);
 const createIterationRecommendationChangeSetDraftMock = vi.mocked(createPMFlowIterationRecommendationChangeSetDraft);
 const createReleaseValidationRunMock = vi.mocked(createPMFlowReleaseValidationRun);
 const createReleaseFeedbackIngestMock = vi.mocked(createPMFlowReleaseFeedbackIngest);
@@ -703,6 +706,107 @@ describe("PMFlowOsProvider real change review actions", () => {
     );
   });
 
+  it("creates a grouped ChangeSet draft from recommendation quality groups", async () => {
+    getExecutionSummaryMock.mockResolvedValue({
+      ...executionSummaryWithGate(releaseGate("ready")),
+      releaseSignals: {
+        status: "active",
+        relationship: {
+          releaseGateStatus: "ready",
+          latestValidationRunId: "900",
+          latestValidationRunStatus: "ready",
+          source: "pmflow_os_release_growth",
+          changeSetCreated: false,
+        },
+        summary: { feedbackCount: 2, revenueEventCount: 0, iterationRecommendationCount: 2, latestSignalAt: "2026-04-26T00:00:00" },
+        feedback: [],
+        revenueEvents: [],
+        recommendations: [
+          {
+            id: "rec-1",
+            projectId: "123",
+            feedbackIngestId: "feedback-1",
+            status: "draft",
+            priority: "high",
+            recommendationSummary: { summary: "补齐失败样本。" },
+            proposedCharterPatch: {},
+            proposedTasks: [],
+            changeSetCreated: false,
+            groupKey: "quality",
+            mergedCount: 2,
+            impactLevel: "high",
+            riskLevel: "high",
+            suggestedChangeTypes: ["Requirement", "Decision", "Task", "Risk"],
+            reviewPriority: "p0",
+          },
+          {
+            id: "rec-2",
+            projectId: "123",
+            feedbackIngestId: "feedback-2",
+            status: "draft",
+            priority: "high",
+            recommendationSummary: { summary: "修复幻觉回答。" },
+            proposedCharterPatch: {},
+            proposedTasks: [],
+            changeSetCreated: false,
+            groupKey: "quality",
+            duplicateOf: "rec-1",
+            mergedCount: 2,
+            impactLevel: "high",
+            riskLevel: "high",
+            suggestedChangeTypes: ["Requirement", "Decision", "Task", "Risk"],
+            reviewPriority: "p0",
+          },
+        ],
+        recommendationGroups: [
+          {
+            groupKey: "quality",
+            title: "quality 反馈组",
+            recommendationIds: ["rec-1", "rec-2"],
+            mergedCount: 2,
+            impactLevel: "high",
+            riskLevel: "high",
+            suggestedChangeTypes: ["Requirement", "Decision", "Task", "Risk"],
+            reviewPriority: "p0",
+            changeSetCreated: false,
+            rationale: "聚焦 quality；负向反馈 4 条；建议任务 2 个。",
+          },
+        ],
+        nextActions: ["同组 recommendation 可批量生成下一轮待确认 ChangeSet draft。"],
+        warnings: [],
+      },
+    });
+    createIterationRecommendationGroupChangeSetDraftMock.mockResolvedValue({
+      changeSetDetail: {
+        changeSet: { id: "88", status: "open" },
+        items: [{ id: "89", objectType: "Requirement", status: "pending_review" }],
+        evidence: [{ id: "90", sourceType: "iteration_recommendation_group", metadata: { groupKey: "quality" } }],
+        approvals: [],
+      },
+      executionSummary: executionSummaryWithGate(releaseGate("ready")),
+      warnings: [],
+    });
+    const { default: ReleaseGrowthPage } = await import("@/features/pmflow-os/pages/ReleaseGrowthPage");
+
+    render(<ReleaseGrowthPage />, { wrapper: wrapper() });
+
+    expect(await screen.findByText("quality 反馈组")).toBeTruthy();
+    expect(screen.getByText("聚焦 quality；负向反馈 4 条；建议任务 2 个。")).toBeTruthy();
+    const action = await screen.findByText("批量生成这一组待确认变更");
+    fireEvent.click(action);
+
+    await waitFor(() =>
+      expect(createIterationRecommendationGroupChangeSetDraftMock).toHaveBeenCalledWith(
+        "123",
+        "quality",
+        expect.objectContaining({
+          recommendation_ids: ["rec-1", "rec-2"],
+          note: expect.stringContaining("quality"),
+        })
+      )
+    );
+  });
+
   it("writes release feedback through the PMFlow OS api in real mode", async () => {
     createReleaseFeedbackIngestMock.mockResolvedValue({
       ingest: {
@@ -771,6 +875,7 @@ describe("PMFlowOsProvider real change review actions", () => {
     expect(createExecutionProgressEventMock).not.toHaveBeenCalled();
     expect(createReleaseValidationRunMock).not.toHaveBeenCalled();
     expect(createReleaseFeedbackIngestMock).not.toHaveBeenCalled();
+    expect(createIterationRecommendationGroupChangeSetDraftMock).not.toHaveBeenCalled();
     expect(createIterationRecommendationChangeSetDraftMock).not.toHaveBeenCalled();
     expect(getExecutionSummaryMock).not.toHaveBeenCalled();
     expect(result.current.state.changeItems.find((item) => item.id === pending!.id)?.status).toBe("accepted");

frontend/src/features/pmflow-os/store.tsx

@@ -7,6 +7,7 @@ import {
   createPMFlowExecutionHandoff,
   createPMFlowExecutionProgressEvent,
   createPMFlowExtraction,
+  createPMFlowIterationRecommendationGroupChangeSetDraft,
   createPMFlowIterationRecommendationChangeSetDraft,
   createPMFlowReleaseFeedbackIngest,
   createPMFlowReleaseValidationRun,
@@ -19,6 +20,7 @@ import {
   type PMFlowExecutionProgressEventRequest,
   type PMFlowExecutionSummary,
   type PMFlowIterationRecommendationChangeSetDraftRequest,
+  type PMFlowIterationRecommendationGroupChangeSetDraftRequest,
   type PMFlowProposedChangeItem,
   type PMFlowReleaseFeedbackIngestRequest,
   type PMFlowReleaseValidationRunRequest,
@@ -549,6 +551,10 @@ interface PMFlowOsStoreValue {
     recommendationId: string,
     payload?: PMFlowIterationRecommendationChangeSetDraftRequest
   ) => void;
+  createIterationRecommendationGroupChangeSetDraft: (
+    groupKey: string,
+    payload?: PMFlowIterationRecommendationGroupChangeSetDraftRequest
+  ) => void;
   advanceExecution: () => void;
   writeBackExecution: (outcome: "passed" | "failed") => void;
   publishRelease: () => void;
@@ -802,6 +808,18 @@ export function PMFlowOsProvider({ children }: { children: ReactNode }) {
           }
           mutateDemoOnly((current) => ingestGrowthSignalMutation(current));
         },
+        createIterationRecommendationGroupChangeSetDraft: (groupKey, payload) => {
+          if (dataSource.mode === "api" && dataSource.projectId) {
+            runApiAction(() =>
+              createPMFlowIterationRecommendationGroupChangeSetDraft(dataSource.projectId || "", groupKey, {
+                recommendation_ids: payload?.recommendation_ids || [],
+                note: payload?.note || "ReleaseGrowthPage 批量生成同组下一轮待确认 ChangeSet draft；不自动 commit。",
+              })
+            );
+            return;
+          }
+          mutateDemoOnly((current) => ingestGrowthSignalMutation(current));
+        },
         advanceExecution: () => mutateDemoOnly((current) => advanceExecutionMutation(current)),
         writeBackExecution: (outcome) => mutateDemoOnly((current) => writeBackExecutionMutation(current, outcome)),
         publishRelease: () => mutateDemoOnly((current) => publishReleaseMutation(current)),