V10.494 add market intel candidate queue review gate

TODO_NEXT_STEPS.txt

@@ -4,6 +4,7 @@
 ================================================================================
 
 【已完成】
+   - V10.494 新增市場情報 MCP Fetch Candidate Queue Review 安全預覽 gate：只審核 candidate handoff 後的人工 queue review 草案，要求候選 key 對齊、review_state 停在 needs_review、allowed actions 限定人工確認/否決/延後、queue_write_status 維持 not_persisted；API 不建立 queue、不更新 review_state、不寫 DB、不連外、不掛 scheduler。
    - V10.493 新增市場情報 MCP Fetch Candidate Handoff Review 安全預覽 gate：只審核 parser review 後的候選交接包，確認 source/candidate key 對齊、queue policy 仍是 manual preview、候選數維持小批次、無 raw/secret/side-effect；API 不建立 queue、不寫 DB、不讀 artifact、不連外、不掛 scheduler。
    - V10.492 收緊 PChome 近門檻自動回刷隊列：`retryable_candidate_revalidation` 不再把 `identity_veto`、`unit_comparable`、`true_low_confidence` 納入每日自動回刷；只處理 `recoverable_low_score` 與 legacy `low_score / refresh_low_score`，並要求無 hard veto、仍在 `exact_identity`、且具備同品線/identity anchor 證據。這讓「可救回」與「正確阻擋」在操作層面真正分流，避免為了壓低 low_score 而重跑不該自動推進的候選。
    - V10.491 新增市場情報 MCP Fetch Result Parser Review 安全預覽 gate：只審核操作員貼回的 parser 結構化摘要，對齊 receipt source/path、候選必要欄位、公開 URL、小批次上限與 raw HTML/secret/side-effect 風險；API 不讀 artifact、不執行 parser CLI、不抓外站、不寫 DB、不掛 scheduler。

config.py

@@ -350,7 +350,7 @@ YOUTUBE_API_KEY = os.getenv('YOUTUBE_API_KEY', '')
 # ==========================================
 # 系統版本與路徑
 # ==========================================
-SYSTEM_VERSION = "V10.493"
+SYSTEM_VERSION = "V10.494"
 LOG_FILE_PATH = os.path.join(BASE_DIR, 'logs/system.log')
 public_url = PUBLIC_URL  # 用於模板顯示
 

docs/adr/ADR-035-cross-platform-market-campaign-intelligence.md

@@ -165,6 +165,7 @@ EwoooC 目前已有 MOMO EDM / 節慶活動資料、`promo_products`、PChome
 - 2026-05-29 追加 MCP fetch run receipt gate：`services.market_intel.mcp_fetch_run_receipt` 與 `/api/market_intel/mcp_fetch_run_receipt` 在操作員 shell 完成 dry-run fetch 後審核 receipt id、artifact path、platform/source/receipt path 對帳、公開 URL、request/error budget、secret 外洩與 API/DB/scheduler 副作用旗標；API/UI 不執行 CLI、不發外部 request、不保存 receipt、不開 DB、不寫入、不掛 scheduler，只放行到後續 result parser review。
 - 2026-05-31 追加 MCP fetch result parser review gate：`services.market_intel.mcp_fetch_result_parser_review` 與 `/api/market_intel/mcp_fetch_result_parser_review` 在 receipt 通過後審核操作員貼回的 parser 結構化摘要，檢查 receipt source/receipt path 對帳、parser artifact path、活動/商品候選必要欄位、公開 URL、小批次候選上限、raw HTML/secret 外洩與 API/DB/scheduler 副作用旗標；API/UI 不讀 artifact、不執行 parser CLI、不發外部 request、不保存 parser result、不寫入、不掛 scheduler，只放行到候選 handoff review。
 - 2026-05-31 追加 MCP fetch candidate handoff review gate：`services.market_intel.mcp_fetch_candidate_handoff_review` 與 `/api/market_intel/mcp_fetch_candidate_handoff_review` 在 parser review 通過後審核候選交接包，檢查 source/candidate key 對齊、queue policy 是否仍是 `manual_candidate_review` / `preview_only`、小批次候選上限、操作員無寫入/無連外/無排程確認、raw HTML/secret 外洩與 side-effect flags；API/UI 不建立 queue、不讀 artifact、不寫 DB、不掛 scheduler，只放行到人工 candidate queue review。
+- 2026-05-31 追加 MCP fetch candidate queue review gate：`services.market_intel.mcp_fetch_candidate_queue_review` 與 `/api/market_intel/mcp_fetch_candidate_queue_review` 在 handoff review 通過後審核人工 queue review 草案，檢查候選 key 對齊、`review_state=needs_review`、allowed actions 人工限定、`queue_write_status=not_persisted`、操作員無寫入/無連外/無排程確認、raw HTML/secret 外洩與 side-effect flags；API/UI 不建立 queue、不更新 review_state、不讀 artifact、不寫 DB、不掛 scheduler，只放行到 writer preflight。
 - 2026-05-18 追加 scheduler attach plan preview：`services.market_intel.scheduler_plan` 與 `/api/market_intel/scheduler_plan` 描述未來 `campaign_discovery_daily`、`campaign_product_probe`、`product_match_review_seed` 三個 job 的 cadence、gate、fallback 與安全邊界。此階段不註冊 scheduler job、不啟動 crawler、不連外、不寫 DB；排程掛載必須等 migration、seed、MCP fetch gate、manual sample 與人工批准全過。
 - 2026-05-18 追加 match review plan preview：`services.market_intel.match_review_plan` 與 `/api/market_intel/match_review_plan` 定義商品比對訊號、分數門檻、`needs_review → confirmed/rejected` HITL 流程與安全邊界。此階段不建立 review queue、不自動 confirmed、不寫 `market_product_matches`、不呼叫 MCP；價格只能作為輔助訊號，不能單獨決定同品比對。
 - 2026-05-18 追加 opportunity plan preview：`services.market_intel.opportunity_plan` 與 `/api/market_intel/opportunity_plan` 定義競品低價威脅、促銷缺口、深折重疊、活動即將結束四類規則與分級策略。此階段不建立 opportunity queue、不派送 Telegram、不產生 AI 摘要、不寫 DB；高風險項必須先有 confirmed match 與 DB evidence 才能升級。

docs/memory/code_modularization_inventory_20260430.md

@@ -42,6 +42,7 @@
 - 2026-05-24 追記：同步市場情報 MCP fetch target review gate 後的 `routes/market_intel_routes.py` 與 `services/market_intel/deployment_readiness.py` 行數；本次新增邏輯已放在獨立 `services/market_intel/mcp_fetch_target_review.py`，route 僅保留薄 glue，後續市場情報 MCP route 應拆出子 Blueprint 或 route registration helper。
 - 2026-05-24 追記：同步市場情報 MCP fetch run package gate 後的 `routes/market_intel_routes.py` 與 `services/market_intel/deployment_readiness.py` 行數；本次新增 endpoint 已拆到 `routes/market_intel_mcp_run_routes.py`，主 Blueprint 只新增 extension import，後續 MCP route 應延續此模式。
 - 2026-05-24 追記：同步市場情報 MCP fetch run readiness gate 後的 `services/market_intel/deployment_readiness.py` 行數；本次新增 endpoint 延續 `routes/market_intel_mcp_run_routes.py` route extension，新增邏輯放在獨立 `services/market_intel/mcp_fetch_run_readiness.py`。
+- 2026-05-31 追記：同步市場情報 MCP fetch candidate queue review gate 後的 `services/market_intel/deployment_readiness.py` 行數；本次新增邏輯維持在獨立 `services/market_intel/mcp_fetch_candidate_queue_review.py`，route 延續 `routes/market_intel_mcp_run_routes.py` extension。
 - 2026-05-24 追記：同步背景 Code Review 111 fallback 保護合併後的 `services/code_review_pipeline_service.py` 行數；此處只更新 inventory，不變更 Code Review 行為。
 - 2026-05-21 追記：同步 PChome/LUDEYA 商品線名稱漂移比對更新後的 `services/marketplace_product_matcher.py` 行數；此處只更新 inventory，不變更模組化決策。
 - 2026-05-21 追記：同步 MAC/Yuskin/AHC 名稱漂移與 bundle equivalent matcher 更新後的 `services/marketplace_product_matcher.py` 行數；此處只更新 inventory，不變更模組化決策。
@@ -94,7 +95,7 @@
 | 805 | `routes/bot_api_routes.py` | P2 Bot API Blueprint | route glue / bot action service |
 | 1319 | `routes/market_intel_review_report_routes.py` | P2 market intel review report Blueprint | review report route glue / export payload / phase handoff orchestration |
 | 917 | `routes/market_intel_routes.py` | P2 market intel Blueprint | page route / API route glue / MCP gate route registration helper |
-| 1040 | `services/market_intel/deployment_readiness.py` | P2 market intel deployment readiness | preflight gates / readiness payload / route contract helpers |
+| 1181 | `services/market_intel/deployment_readiness.py` | P2 market intel deployment readiness | preflight gates / readiness payload / route contract helpers |
 | 846 | `services/market_intel/candidate_queue_review_ai_summary_persistence_telegram_dispatch_report_catalog_record_run_receipt.py` | P2 market intel review receipt pipeline | AI summary / persistence / Telegram dispatch / report catalog run receipt orchestration |
 
 ## 市場情報開發前置禁區

docs/memory/current_execution_queue_20260524.md

@@ -89,6 +89,7 @@
 - 2026-05-31 起，`V10.491` 新增市場情報 MCP Fetch Result Parser Review gate：在 receipt gate 後只審核操作員 shell parser 貼回的結構化摘要，對齊 source/path、公開 URL、候選必要欄位、小批次上限、raw HTML/secret/side-effect 風險；仍不讀 artifact、不執行 CLI、不連外、不寫 DB、不掛 scheduler。
 - 2026-05-31 起，`V10.492` 收緊 PChome 近門檻自動回刷：`run_retryable_candidate_revalidation()` 只回刷 `recoverable_low_score` 與 legacy `low_score / refresh_low_score`，且 SQL 端要求 `hard_veto=false`、`comparison_mode=exact_identity`、diagnostic reasons 命中同品線/identity anchor；`identity_veto`、`unit_comparable`、`true_low_confidence` 不再進每日自動回刷隊列，需等待新證據或人工處理。
 - 2026-05-31 起，`V10.493` 新增市場情報 MCP Fetch Candidate Handoff Review gate：在 parser review 通過後只審核候選交接包，要求 source/candidate key 完全對齊、queue policy 維持 manual preview、小批次上限與操作員無寫入/無連外/無排程確認；仍不建立 queue、不寫 DB、不讀 artifact、不連外、不掛 scheduler。
+- 2026-05-31 起，`V10.494` 新增市場情報 MCP Fetch Candidate Queue Review gate：在 handoff review 通過後只審核人工 queue review 草案，要求候選 key 完全對齊、review_state 只停在 `needs_review`、allowed actions 限人工操作、queue_write_status 維持 `not_persisted`；仍不建立 queue、不更新 review_state、不寫 DB、不連外、不掛 scheduler。
 
 ## 3. 12 Agent 決策信封整合
 

docs/memory/history_logs.md

@@ -13,6 +13,7 @@
 ## 📅 詳細更新日誌 (考古存檔)
 
 ### 2026-05-24：PChome 近門檻身份回收第二輪
+- **V10.494 市場情報 MCP Fetch Candidate Queue Review gate**: 新增 `/api/market_intel/mcp_fetch_candidate_queue_review` 與 UI preview，只審核 candidate handoff 後的人工 queue review 草案；要求候選 key 完全對齊、`review_state=needs_review`、allowed actions 限人工確認/否決/延後、`queue_write_status=not_persisted`，且 API 不建立 queue、不更新 review_state、不寫 DB、不連外、不掛 scheduler。
 - **V10.493 市場情報 MCP Fetch Candidate Handoff Review gate**: 新增 `/api/market_intel/mcp_fetch_candidate_handoff_review` 與 UI preview，只審核 parser review 後的候選交接包；要求 source/candidate key 完全對齊、queue policy 維持 `manual_candidate_review` / `preview_only`、候選數維持小批次，且 API 不建立 queue、不寫 DB、不讀 artifact、不連外、不掛 scheduler。
 - **V10.491 市場情報 MCP Fetch Result Parser Review gate**: 新增 `/api/market_intel/mcp_fetch_result_parser_review` 與 UI preview，只審核操作員 shell parser 後貼回的結構化摘要；API 不讀 artifact、不執行 parser CLI、不抓外站、不寫檔、不開 DB、不掛 scheduler，且會阻擋 raw HTML/body/snapshot、secret/token 欄位與 side-effect flags。
 - **V10.489 PChome 低分同款人工覆核回收與 gate-pass 風險邊界**: `marketplace_product_matcher` 新增三個窄範圍 focused identity：TS6 超美白香氛誘霜 120g/ml、W 修護保養蝸牛特潤修護面膜 6 片、Derma 大地 Eco 植萃護膚油 2 入。這些樣本只升到 `identity_review / manual_review`，不進 `price_alert_exact`；同版補 Clarins 身體油不同線、命名組合品數量反轉、isLeaf 香型數量不一致 hard veto，HOOOME 大理石暖燈單側設計差留人工覆核。

routes/market_intel_mcp_run_routes.py

@@ -22,6 +22,9 @@ from services.market_intel.mcp_fetch_result_parser_review import (
 from services.market_intel.mcp_fetch_candidate_handoff_review import (
     build_mcp_fetch_candidate_handoff_review_preview,
 )
+from services.market_intel.mcp_fetch_candidate_queue_review import (
+    build_mcp_fetch_candidate_queue_review_preview,
+)
 
 
 @market_intel_bp.route("/api/market_intel/mcp_fetch_run_package", methods=["GET", "POST"])
@@ -204,3 +207,46 @@ def market_intel_mcp_fetch_candidate_handoff_review():
             phase=service.phase,
         )
     )
+
+
+@market_intel_bp.route(
+    "/api/market_intel/mcp_fetch_candidate_queue_review",
+    methods=["GET", "POST"],
+)
+@login_required
+def market_intel_mcp_fetch_candidate_queue_review():
+    handoff_review_package = {}
+    handoff_review_result = None
+    candidate_queue_review = None
+    if request.method == "POST":
+        payload = request.get_json(silent=True) or {}
+        package = (
+            payload.get("candidate_queue_review")
+            or payload.get("queue_review")
+            or payload
+        )
+        handoff_review_package = (
+            package.get("handoff_review_package")
+            or package.get("candidate_handoff_review")
+            or package.get("handoff_review")
+            or {}
+        )
+        handoff_review_result = (
+            package.get("handoff_review_result")
+            or package.get("mcp_fetch_candidate_handoff_review")
+        )
+        candidate_queue_review = (
+            package.get("candidate_queue_review")
+            or package.get("queue_review")
+            or package.get("review_payload")
+        )
+
+    service = MarketIntelService()
+    return jsonify(
+        build_mcp_fetch_candidate_queue_review_preview(
+            handoff_review_package=handoff_review_package,
+            handoff_review_result=handoff_review_result,
+            candidate_queue_review=candidate_queue_review,
+            phase=service.phase,
+        )
+    )

services/market_intel/deployment_readiness.py

@@ -78,6 +78,9 @@ from services.market_intel.mcp_fetch_result_parser_review import (
 from services.market_intel.mcp_fetch_candidate_handoff_review import (
     build_mcp_fetch_candidate_handoff_review_preview,
 )
+from services.market_intel.mcp_fetch_candidate_queue_review import (
+    build_mcp_fetch_candidate_queue_review_preview,
+)
 from services.market_intel.mcp_manual_fetch_handoff import (
     build_mcp_manual_fetch_handoff_preview,
 )
@@ -205,6 +208,11 @@ PRODUCTION_SMOKE_TARGETS = (
     + ("/api/market_intel/mcp_fetch_candidate_handoff_review",)
     + PRODUCTION_SMOKE_TARGETS[-1:]
 )
+PRODUCTION_SMOKE_TARGETS = (
+    PRODUCTION_SMOKE_TARGETS[:-1]
+    + ("/api/market_intel/mcp_fetch_candidate_queue_review",)
+    + PRODUCTION_SMOKE_TARGETS[-1:]
+)
 def _run_review_preview_safe(payload, mode):
     return bool(payload["mode"] == mode and all(not payload.get(key) for key in BLOCKED_RUN_REVIEW_KEYS))
 def build_deployment_readiness_preview(*, service, market_intel_tables, schema_smoke_builder):
@@ -249,6 +257,11 @@ def build_deployment_readiness_preview(*, service, market_intel_tables, schema_s
             phase=service.phase,
         )
     )
+    mcp_fetch_candidate_queue_review = (
+        build_mcp_fetch_candidate_queue_review_preview(
+            phase=service.phase,
+        )
+    )
     scheduler_plan = service.build_scheduler_plan()
     manual_sample_plan = service.build_manual_sample_plan()
     manual_sample_acceptance = service.build_manual_sample_acceptance()
@@ -627,6 +640,28 @@ def build_deployment_readiness_preview(*, service, market_intel_tables, schema_s
             and not mcp_fetch_candidate_handoff_review["file_written"]
             and not mcp_fetch_candidate_handoff_review["scheduler_attached"]
         ),
+        "mcp_fetch_candidate_queue_review_preview_safe": bool(
+            mcp_fetch_candidate_queue_review["mode"]
+            == "mcp_fetch_candidate_queue_review_preview"
+            and not mcp_fetch_candidate_queue_review["payload_persisted"]
+            and not mcp_fetch_candidate_queue_review["candidate_queue_created"]
+            and not mcp_fetch_candidate_queue_review["candidate_queue_persisted"]
+            and not mcp_fetch_candidate_queue_review[
+                "candidate_review_state_updated"
+            ]
+            and not mcp_fetch_candidate_queue_review["network_request_allowed"]
+            and not mcp_fetch_candidate_queue_review["api_executes_cli"]
+            and not mcp_fetch_candidate_queue_review[
+                "api_opens_database_connection"
+            ]
+            and not mcp_fetch_candidate_queue_review["api_writes_database"]
+            and not mcp_fetch_candidate_queue_review["api_uses_external_network"]
+            and not mcp_fetch_candidate_queue_review["database_write_executed"]
+            and not mcp_fetch_candidate_queue_review["fetch_executed"]
+            and not mcp_fetch_candidate_queue_review["cli_executed"]
+            and not mcp_fetch_candidate_queue_review["file_written"]
+            and not mcp_fetch_candidate_queue_review["scheduler_attached"]
+        ),
         "scheduler_plan_preview_safe": bool(
             scheduler_plan["mode"] == "scheduler_attach_plan_preview"
             and not scheduler_plan["scheduler_registration_executed"]
@@ -1068,6 +1103,7 @@ def build_deployment_readiness_preview(*, service, market_intel_tables, schema_s
         "mcp_fetch_run_receipt": mcp_fetch_run_receipt,
         "mcp_fetch_result_parser_review": mcp_fetch_result_parser_review,
         "mcp_fetch_candidate_handoff_review": mcp_fetch_candidate_handoff_review,
+        "mcp_fetch_candidate_queue_review": mcp_fetch_candidate_queue_review,
         "scheduler_plan": scheduler_plan,
         "manual_sample_plan": manual_sample_plan,
         "manual_sample_acceptance": manual_sample_acceptance,

services/market_intel/mcp_fetch_candidate_queue_review.py

@@ -0,0 +1,636 @@
+"""市場情報 MCP fetch candidate queue review preview。
+
+本模組只審核 candidate handoff 後的人工 queue review 草案；
+不建立 queue、不寫 DB、不讀 artifact、不發 HTTP request、不掛 scheduler。
+"""
+
+from services.market_intel.mcp_fetch_candidate_handoff_review import (
+    build_mcp_fetch_candidate_handoff_review_preview,
+)
+
+
+MAX_QUEUE_REVIEW_ITEMS = 80
+SAFE_REVIEW_STATES = {"needs_review"}
+SAFE_REVIEW_MODES = {"operator_queue_review_preview"}
+SAFE_TARGET_QUEUES = {"manual_candidate_review"}
+SAFE_ALLOWED_ACTIONS = {
+    "confirm_campaign_candidate",
+    "confirm_product_candidate",
+    "reject_candidate",
+    "defer_candidate",
+}
+
+FORBIDDEN_SECRET_KEYS = (
+    "approval_token",
+    "approval-token",
+    "api_key",
+    "authorization",
+    "bearer",
+    "client_secret",
+    "cookie",
+    "password",
+    "secret",
+    "session_cookie",
+    "token",
+)
+
+SAFE_SECRET_METADATA_KEYS = {
+    "operator_confirmed_no_secret_payload",
+}
+
+FORBIDDEN_RAW_PAYLOAD_KEYS = (
+    "body_html",
+    "document_html",
+    "html",
+    "page_html",
+    "raw_body",
+    "raw_html",
+    "raw_snapshot",
+    "response_body",
+)
+
+_BLOCKED_SIDE_EFFECT_KEYS = (
+    "allow_api_execution",
+    "allow_database_write",
+    "allow_external_network_in_api",
+    "allow_scheduler_attach",
+    "api_executed_cli",
+    "api_executes_cli",
+    "api_executes_docker",
+    "api_executes_health_check",
+    "api_executes_ssh",
+    "api_opens_database_connection",
+    "api_uses_external_network",
+    "api_writes_database",
+    "attach_scheduler",
+    "candidate_queue_created",
+    "candidate_queue_persisted",
+    "candidate_review_state_updated",
+    "cli_executed",
+    "command_executed",
+    "database_commit_executed",
+    "database_session_created",
+    "database_write_executed",
+    "external_network_executed",
+    "fetch_executed",
+    "file_written",
+    "network_request_allowed",
+    "payload_persisted",
+    "scheduler_attached",
+    "write_database",
+    "writes_executed",
+    "would_write_database",
+)
+
+
+def _as_dict(value):
+    return value if isinstance(value, dict) else {}
+
+
+def _as_list(value):
+    if value is None:
+        return []
+    if isinstance(value, (list, tuple, set)):
+        return list(value)
+    return [value]
+
+
+def _safe_int(value):
+    try:
+        return int(value or 0)
+    except (TypeError, ValueError):
+        return 0
+
+
+def _has_text(value):
+    return bool(isinstance(value, str) and value.strip())
+
+
+def _safe_text(value, limit=500):
+    if value is None:
+        return None
+    text = str(value).strip()
+    return text[:limit] if text else None
+
+
+def _contains_forbidden_key(value, forbidden_keys, *, safe_keys=None):
+    safe_keys = safe_keys or set()
+    if isinstance(value, dict):
+        for key, nested in value.items():
+            normalized_key = str(key).lower()
+            if normalized_key in safe_keys and isinstance(nested, bool):
+                continue
+            if any(forbidden_key in normalized_key for forbidden_key in forbidden_keys):
+                return True
+            if _contains_forbidden_key(nested, forbidden_keys, safe_keys=safe_keys):
+                return True
+    elif isinstance(value, list):
+        return any(
+            _contains_forbidden_key(item, forbidden_keys, safe_keys=safe_keys)
+            for item in value
+        )
+    return False
+
+
+def _blocked_side_effects(payload):
+    found = []
+
+    def visit(value, path):
+        if isinstance(value, dict):
+            for key, item in value.items():
+                normalized_key = str(key).lower()
+                key_path = f"{path}.{key}" if path else key
+                if normalized_key in _BLOCKED_SIDE_EFFECT_KEYS and bool(item):
+                    found.append(key_path)
+                visit(item, key_path)
+        elif isinstance(value, list):
+            for index, item in enumerate(value):
+                visit(item, f"{path}[{index}]")
+
+    visit(payload, "")
+    return found
+
+
+def _handoff_review_from_inputs(handoff_review_package, handoff_review_result, phase):
+    if isinstance(handoff_review_result, dict) and handoff_review_result:
+        return handoff_review_result
+
+    handoff_review_package = _as_dict(handoff_review_package)
+    return build_mcp_fetch_candidate_handoff_review_preview(
+        parser_review_package=(
+            handoff_review_package.get("parser_review_package")
+            or handoff_review_package.get("parser_review")
+            or handoff_review_package.get("result_parser_review")
+        ),
+        parser_review_result=(
+            handoff_review_package.get("parser_review_result")
+            or handoff_review_package.get("mcp_fetch_result_parser_review")
+        ),
+        candidate_handoff=(
+            handoff_review_package.get("candidate_handoff")
+            or handoff_review_package.get("handoff")
+            or handoff_review_package.get("handoff_payload")
+        ),
+        phase=phase,
+    )
+
+
+def _candidate_key(candidate, candidate_type):
+    candidate = _as_dict(candidate)
+    if candidate_type == "campaign":
+        return candidate.get("campaign_key")
+    return candidate.get("platform_product_id")
+
+
+def _sample_candidate_queue_review_package():
+    handoff_preview = build_mcp_fetch_candidate_handoff_review_preview()
+    handoff_review_package = handoff_preview["sample_candidate_handoff_package"]
+    handoff_review_result = build_mcp_fetch_candidate_handoff_review_preview(
+        parser_review_package=handoff_review_package["parser_review_package"],
+        parser_review_result=handoff_review_package["parser_review_result"],
+        candidate_handoff=handoff_review_package["candidate_handoff"],
+    )
+    handoff_summary = handoff_review_result["candidate_handoff_summary"]
+    review_items = []
+    for candidate_type, candidates in (
+        ("campaign", handoff_summary.get("campaign_candidates", [])),
+        ("product", handoff_summary.get("product_candidates", [])),
+    ):
+        for item in candidates:
+            item = _as_dict(item)
+            candidate_key = _candidate_key(item, candidate_type)
+            review_items.append(
+                {
+                    "candidate_type": candidate_type,
+                    "platform_code": item.get("platform_code"),
+                    "source_key": item.get("source_key"),
+                    "candidate_key": candidate_key,
+                    "candidate_name": item.get("campaign_name") or item.get("name"),
+                    "candidate_url": item.get("campaign_url") or item.get("product_url"),
+                    "review_state": "needs_review",
+                    "priority_lane": "normal",
+                    "allowed_actions": sorted(SAFE_ALLOWED_ACTIONS),
+                    "evidence_ref": "parser_summary",
+                    "queue_write_status": "not_persisted",
+                }
+            )
+
+    return {
+        "handoff_review_package": handoff_review_package,
+        "handoff_review_result": handoff_review_result,
+        "candidate_queue_review": {
+            "review_id": "market-intel-candidate-queue-review-sample",
+            "review_mode": "operator_queue_review_preview",
+            "target_queue": "manual_candidate_review",
+            "operator_confirmed_no_database_write": True,
+            "operator_confirmed_no_external_network": True,
+            "operator_confirmed_no_scheduler_attach": True,
+            "operator_confirmed_no_persistence": True,
+            "operator_confirmed_manual_review_required": True,
+            "operator_confirmed_no_secret_payload": True,
+            "summary": {
+                "campaign_candidate_count": handoff_summary.get(
+                    "campaign_candidate_count",
+                    0,
+                ),
+                "product_candidate_count": handoff_summary.get(
+                    "product_candidate_count",
+                    0,
+                ),
+                "review_item_count": len(review_items),
+            },
+            "review_items": review_items,
+        },
+    }
+
+
+def _handoff_summary(handoff_review):
+    handoff_review = _as_dict(handoff_review)
+    side_effects_clear = bool(
+        not handoff_review.get("manual_fetch_gate_opened_by_api")
+        and not handoff_review.get("network_request_allowed")
+        and not handoff_review.get("api_executes_cli")
+        and not handoff_review.get("api_opens_database_connection")
+        and not handoff_review.get("api_writes_database")
+        and not handoff_review.get("api_uses_external_network")
+        and not handoff_review.get("database_write_executed")
+        and not handoff_review.get("fetch_executed")
+        and not handoff_review.get("cli_executed")
+        and not handoff_review.get("file_written")
+        and not handoff_review.get("scheduler_attached")
+        and not handoff_review.get("candidate_queue_created")
+        and not handoff_review.get("candidate_queue_persisted")
+    )
+    summary = _as_dict(handoff_review.get("candidate_handoff_summary"))
+    return {
+        "mode": handoff_review.get("mode"),
+        "accepted": bool(
+            handoff_review.get("mcp_fetch_candidate_handoff_review_accepted")
+        ),
+        "ready_for_manual_candidate_queue_review": bool(
+            handoff_review.get("ready_for_manual_candidate_queue_review")
+        ),
+        "source_count": _safe_int(handoff_review.get("handoff_source_count")),
+        "campaign_candidate_count": _safe_int(
+            handoff_review.get("campaign_candidate_count")
+        ),
+        "product_candidate_count": _safe_int(
+            handoff_review.get("product_candidate_count")
+        ),
+        "candidate_count": _safe_int(handoff_review.get("candidate_count")),
+        "source_keys": sorted(
+            tuple(item)
+            for item in _as_list(summary.get("source_keys"))
+            if isinstance(item, (list, tuple)) and len(item) == 2
+        ),
+        "campaign_candidate_keys": sorted(
+            key for key in _as_list(summary.get("campaign_candidate_keys")) if key
+        ),
+        "product_candidate_keys": sorted(
+            key for key in _as_list(summary.get("product_candidate_keys")) if key
+        ),
+        "side_effects_clear": side_effects_clear,
+        "blocked_reasons": handoff_review.get("blocked_reasons", []),
+    }
+
+
+def _review_item_summary(item):
+    item = _as_dict(item)
+    allowed_actions = sorted(
+        action for action in _as_list(item.get("allowed_actions")) if action
+    )
+    return {
+        "candidate_type": _safe_text(item.get("candidate_type"), 40),
+        "platform_code": _safe_text(item.get("platform_code"), 80),
+        "source_key": _safe_text(item.get("source_key"), 160),
+        "candidate_key": _safe_text(item.get("candidate_key"), 240),
+        "candidate_name": _safe_text(item.get("candidate_name"), 300),
+        "candidate_url": _safe_text(item.get("candidate_url"), 500),
+        "review_state": _safe_text(item.get("review_state"), 80),
+        "priority_lane": _safe_text(item.get("priority_lane"), 80),
+        "allowed_actions": allowed_actions,
+        "evidence_ref": _safe_text(item.get("evidence_ref"), 160),
+        "queue_write_status": _safe_text(item.get("queue_write_status"), 80),
+        "required_fields_present": bool(
+            item.get("candidate_type") in {"campaign", "product"}
+            and _has_text(item.get("platform_code"))
+            and _has_text(item.get("source_key"))
+            and _has_text(item.get("candidate_key"))
+            and _has_text(item.get("candidate_name"))
+            and _has_text(item.get("candidate_url"))
+        ),
+        "allowed_actions_safe": bool(
+            allowed_actions
+            and set(allowed_actions).issubset(SAFE_ALLOWED_ACTIONS)
+        ),
+    }
+
+
+def _queue_review_summary(candidate_queue_review):
+    candidate_queue_review = _as_dict(candidate_queue_review)
+    items = [
+        _review_item_summary(item)
+        for item in _as_list(candidate_queue_review.get("review_items"))
+    ]
+    campaign_items = [item for item in items if item["candidate_type"] == "campaign"]
+    product_items = [item for item in items if item["candidate_type"] == "product"]
+    summary = _as_dict(candidate_queue_review.get("summary"))
+    return {
+        "provided_keys": sorted(candidate_queue_review.keys()),
+        "review_id": _safe_text(candidate_queue_review.get("review_id"), 160),
+        "review_mode": _safe_text(candidate_queue_review.get("review_mode"), 120),
+        "target_queue": _safe_text(candidate_queue_review.get("target_queue"), 160),
+        "operator_confirmed_no_database_write": bool(
+            candidate_queue_review.get("operator_confirmed_no_database_write")
+        ),
+        "operator_confirmed_no_external_network": bool(
+            candidate_queue_review.get("operator_confirmed_no_external_network")
+        ),
+        "operator_confirmed_no_scheduler_attach": bool(
+            candidate_queue_review.get("operator_confirmed_no_scheduler_attach")
+        ),
+        "operator_confirmed_no_persistence": bool(
+            candidate_queue_review.get("operator_confirmed_no_persistence")
+        ),
+        "operator_confirmed_manual_review_required": bool(
+            candidate_queue_review.get("operator_confirmed_manual_review_required")
+        ),
+        "operator_confirmed_no_secret_payload": bool(
+            candidate_queue_review.get("operator_confirmed_no_secret_payload")
+        ),
+        "review_item_count": len(items),
+        "summary_review_item_count": _safe_int(summary.get("review_item_count")),
+        "campaign_candidate_count": len(campaign_items),
+        "summary_campaign_candidate_count": _safe_int(
+            summary.get("campaign_candidate_count")
+        ),
+        "product_candidate_count": len(product_items),
+        "summary_product_candidate_count": _safe_int(
+            summary.get("product_candidate_count")
+        ),
+        "campaign_candidate_keys": sorted(
+            item["candidate_key"] for item in campaign_items if item["candidate_key"]
+        ),
+        "product_candidate_keys": sorted(
+            item["candidate_key"] for item in product_items if item["candidate_key"]
+        ),
+        "review_items": items,
+        "all_review_states_safe": bool(
+            items and all(item["review_state"] in SAFE_REVIEW_STATES for item in items)
+        ),
+        "all_required_fields_present": bool(
+            items and all(item["required_fields_present"] for item in items)
+        ),
+        "all_allowed_actions_safe": bool(
+            items and all(item["allowed_actions_safe"] for item in items)
+        ),
+        "all_queue_write_status_preview": bool(
+            items and all(item["queue_write_status"] == "not_persisted" for item in items)
+        ),
+        "summary_counts_match_items": bool(
+            _safe_int(summary.get("review_item_count")) == len(items)
+            and _safe_int(summary.get("campaign_candidate_count")) == len(campaign_items)
+            and _safe_int(summary.get("product_candidate_count")) == len(product_items)
+        ),
+        "raw_payload_submitted_to_api": _contains_forbidden_key(
+            candidate_queue_review,
+            FORBIDDEN_RAW_PAYLOAD_KEYS,
+        ),
+        "secret_or_token_submitted_to_api": _contains_forbidden_key(
+            candidate_queue_review,
+            FORBIDDEN_SECRET_KEYS,
+            safe_keys=SAFE_SECRET_METADATA_KEYS,
+        ),
+        "blocked_side_effects": _blocked_side_effects(candidate_queue_review),
+    }
+
+
+def _queue_review_gates(
+    *,
+    handoff_review_received,
+    queue_review_payload_received,
+    queue_review_valid_object,
+    handoff,
+    queue_review,
+):
+    handoff_campaign_keys = set(handoff["campaign_candidate_keys"])
+    queue_campaign_keys = set(queue_review["campaign_candidate_keys"])
+    handoff_product_keys = set(handoff["product_candidate_keys"])
+    queue_product_keys = set(queue_review["product_candidate_keys"])
+    operator_boundaries_confirmed = bool(
+        queue_review["operator_confirmed_no_database_write"]
+        and queue_review["operator_confirmed_no_external_network"]
+        and queue_review["operator_confirmed_no_scheduler_attach"]
+        and queue_review["operator_confirmed_no_persistence"]
+        and queue_review["operator_confirmed_manual_review_required"]
+        and queue_review["operator_confirmed_no_secret_payload"]
+    )
+    return [
+        {
+            "key": "handoff_review_payload_or_result_received",
+            "label": "已提供 candidate handoff review package 或已審核結果",
+            "passed": handoff_review_received,
+        },
+        {
+            "key": "handoff_review_accepted",
+            "label": "candidate handoff review gate 必須已通過",
+            "passed": handoff["accepted"],
+        },
+        {
+            "key": "handoff_ready_for_queue_review",
+            "label": "handoff review 只放行到人工 candidate queue review",
+            "passed": handoff["ready_for_manual_candidate_queue_review"],
+        },
+        {
+            "key": "handoff_side_effect_free",
+            "label": "handoff review 未顯示 API 執行、連外、寫檔、寫 DB 或掛 scheduler",
+            "passed": handoff["side_effects_clear"],
+        },
+        {
+            "key": "queue_review_payload_received",
+            "label": "已提供 candidate queue review 草案",
+            "passed": queue_review_payload_received,
+        },
+        {
+            "key": "queue_review_valid_object",
+            "label": "candidate queue review payload 必須是 JSON object",
+            "passed": queue_review_valid_object,
+        },
+        {
+            "key": "queue_review_identity_recorded",
+            "label": "candidate queue review 必須記錄 review_id",
+            "passed": bool(queue_review["review_id"]),
+        },
+        {
+            "key": "queue_review_mode_preview_only",
+            "label": "candidate queue review 必須維持 operator_queue_review_preview",
+            "passed": queue_review["review_mode"] in SAFE_REVIEW_MODES,
+        },
+        {
+            "key": "queue_review_target_manual_only",
+            "label": "target queue 只能是 manual_candidate_review",
+            "passed": queue_review["target_queue"] in SAFE_TARGET_QUEUES,
+        },
+        {
+            "key": "queue_review_candidates_match_handoff",
+            "label": "queue review 候選 key 必須完全對齊 handoff review",
+            "passed": bool(
+                handoff_campaign_keys
+                and handoff_product_keys
+                and queue_campaign_keys == handoff_campaign_keys
+                and queue_product_keys == handoff_product_keys
+            ),
+        },
+        {
+            "key": "queue_review_item_count_within_limit",
+            "label": "queue review item 數需維持小批次上限",
+            "passed": bool(0 < queue_review["review_item_count"] <= MAX_QUEUE_REVIEW_ITEMS),
+        },
+        {
+            "key": "queue_review_counts_match_summary",
+            "label": "summary count 必須與 review_items 數量一致",
+            "passed": queue_review["summary_counts_match_items"],
+        },
+        {
+            "key": "queue_review_required_fields_present",
+            "label": "每個 review item 必須具備候選識別欄位",
+            "passed": queue_review["all_required_fields_present"],
+        },
+        {
+            "key": "queue_review_state_needs_review_only",
+            "label": "review_state 必須停在 needs_review",
+            "passed": queue_review["all_review_states_safe"],
+        },
+        {
+            "key": "queue_review_allowed_actions_safe",
+            "label": "allowed actions 必須限定人工確認/否決/延後",
+            "passed": queue_review["all_allowed_actions_safe"],
+        },
+        {
+            "key": "queue_review_not_persisted",
+            "label": "queue_write_status 必須是 not_persisted",
+            "passed": queue_review["all_queue_write_status_preview"],
+        },
+        {
+            "key": "queue_review_operator_boundaries_confirmed",
+            "label": "操作員確認無寫入、無連外、無排程、無保存且需人工覆核",
+            "passed": operator_boundaries_confirmed,
+        },
+        {
+            "key": "queue_review_no_raw_payload",
+            "label": "queue review payload 不得回貼 raw HTML/body/snapshot",
+            "passed": not queue_review["raw_payload_submitted_to_api"],
+        },
+        {
+            "key": "queue_review_no_secret_or_token_key",
+            "label": "queue review payload 不得包含 secret、cookie、password 或 token key",
+            "passed": not queue_review["secret_or_token_submitted_to_api"],
+        },
+        {
+            "key": "queue_review_side_effect_free",
+            "label": "queue review payload 不得要求 API 執行、連外、寫檔、寫 DB 或掛 scheduler",
+            "passed": not queue_review["blocked_side_effects"],
+        },
+    ]
+
+
+def build_mcp_fetch_candidate_queue_review_preview(
+    *,
+    handoff_review_package=None,
+    handoff_review_result=None,
+    candidate_queue_review=None,
+    phase=None,
+):
+    """建立 fetch candidate queue review；不建立 queue 或寫入。"""
+    handoff_review_package = _as_dict(handoff_review_package)
+    handoff_review_result_received = bool(
+        isinstance(handoff_review_result, dict) and handoff_review_result
+    )
+    queue_review_valid_object = isinstance(candidate_queue_review, dict) if (
+        candidate_queue_review is not None
+    ) else True
+    queue_review_payload = _as_dict(candidate_queue_review)
+    handoff_review = _handoff_review_from_inputs(
+        handoff_review_package,
+        handoff_review_result,
+        phase,
+    )
+    handoff_review_received = bool(
+        handoff_review_package or handoff_review_result_received
+    )
+    payload_received = bool(
+        handoff_review_received
+        or queue_review_payload
+        or candidate_queue_review is not None
+    )
+    queue_review_payload_received = bool(queue_review_payload)
+    handoff = _handoff_summary(handoff_review)
+    queue_review = _queue_review_summary(queue_review_payload)
+    gates = _queue_review_gates(
+        handoff_review_received=handoff_review_received,
+        queue_review_payload_received=queue_review_payload_received,
+        queue_review_valid_object=queue_review_valid_object,
+        handoff=handoff,
+        queue_review=queue_review,
+    )
+    blocked_reasons = [gate["key"] for gate in gates if not gate["passed"]]
+    accepted = bool(payload_received and not blocked_reasons)
+
+    return {
+        "mode": (
+            "mcp_fetch_candidate_queue_review"
+            if payload_received
+            else "mcp_fetch_candidate_queue_review_preview"
+        ),
+        "phase": phase,
+        "candidate_queue_review_payload_received": payload_received,
+        "handoff_review_received": handoff_review_received,
+        "queue_review_payload_received": queue_review_payload_received,
+        "queue_review_valid_object": queue_review_valid_object,
+        "handoff_review_accepted": handoff["accepted"],
+        "mcp_fetch_candidate_queue_review_accepted": accepted,
+        "candidate_queue_review_ready": accepted,
+        "ready_for_candidate_queue_writer_preflight": accepted,
+        "ready_for_api_database_write": False,
+        "ready_for_scheduler_attach": False,
+        "network_request_allowed": False,
+        "fetch_executed": False,
+        "cli_executed": False,
+        "candidate_queue_created": False,
+        "candidate_queue_persisted": False,
+        "candidate_review_state_updated": False,
+        "handoff_source_count": handoff["source_count"],
+        "review_item_count": queue_review["review_item_count"],
+        "campaign_candidate_count": queue_review["campaign_candidate_count"],
+        "product_candidate_count": queue_review["product_candidate_count"],
+        "gate_count": len(gates),
+        "passed_gate_count": sum(1 for gate in gates if gate["passed"]),
+        "blocked_reasons": blocked_reasons,
+        "gates": gates,
+        "handoff_review_summary": handoff,
+        "candidate_queue_review_summary": queue_review,
+        "sample_candidate_queue_review_package": (
+            _sample_candidate_queue_review_package()
+        ),
+        "next_operator_steps": [
+            "queue review 通過後，只代表可進 writer preflight，不代表可寫 market_*",
+            "candidate queue writer、DB import、scheduler attach、AI/Telegram 摘要都必須另開獨立 gate",
+            "API/UI 仍不得建立 queue、不得讀 artifact、不得抓外站、不得寫 DB、不得掛 scheduler",
+        ],
+        "payload_persisted": False,
+        "api_executes_health_check": False,
+        "api_executes_docker": False,
+        "api_executes_ssh": False,
+        "api_executes_cli": False,
+        "api_opens_database_connection": False,
+        "api_writes_database": False,
+        "api_uses_external_network": False,
+        "database_session_created": False,
+        "database_commit_executed": False,
+        "database_write_executed": False,
+        "external_network_executed": False,
+        "file_written": False,
+        "writes_executed": False,
+        "would_write_database": False,
+        "scheduler_attached": False,
+    }

services/market_intel/phase.py

@@ -1,3 +1,3 @@
 """市場情報 rollout phase 單一來源。"""
 
-MARKET_INTEL_PHASE = "phase_126_market_intel_mcp_fetch_candidate_handoff_review"
+MARKET_INTEL_PHASE = "phase_127_market_intel_mcp_fetch_candidate_queue_review"

templates/market_intel/disabled.html

@@ -812,6 +812,32 @@
         </div>
     </div>
 
+    <div class="market-intel-panel" data-market-intel-mcp-fetch-candidate-queue-review>
+        <div class="market-intel-preview-head">
+            <div>
+                <p class="market-intel-muted momo-mono mb-1">MCP / FETCH CANDIDATE QUEUE</p>
+                <h2 class="market-intel-preview-title">MCP Fetch Candidate Queue 審核</h2>
+            </div>
+            <button class="market-intel-icon-button" type="button" title="重新整理 MCP Fetch Candidate Queue" data-market-intel-mcp-fetch-candidate-queue-review-refresh>
+                <i class="fas fa-rotate-right" aria-hidden="true"></i>
+            </button>
+        </div>
+        <div class="market-intel-preview-meta" data-market-intel-mcp-fetch-candidate-queue-review-meta>
+            <span class="market-intel-pill">loading</span>
+        </div>
+        <div data-market-intel-mcp-fetch-candidate-queue-review-body>
+            <div class="market-intel-empty">讀取 MCP Fetch Candidate Queue 審核中...</div>
+        </div>
+        <div class="market-intel-control-row mt-3">
+            <textarea class="market-intel-json-input" rows="9" spellcheck="false" data-market-intel-mcp-fetch-candidate-queue-review-input placeholder="handoff review and candidate queue review JSON"></textarea>
+            <div class="market-intel-control-actions">
+                <button class="market-intel-icon-button" type="button" title="審核 MCP Fetch Candidate Queue JSON" data-market-intel-mcp-fetch-candidate-queue-review-review>
+                    <i class="fas fa-check" aria-hidden="true"></i>
+                </button>
+            </div>
+        </div>
+    </div>
+
     <div class="market-intel-panel" data-market-intel-manual-sample>
         <div class="market-intel-preview-head">
             <div>
@@ -1326,6 +1352,7 @@
         const mcpFetchRunReceiptRoot = document.querySelector('[data-market-intel-mcp-fetch-run-receipt]');
         const mcpFetchResultParserReviewRoot = document.querySelector('[data-market-intel-mcp-fetch-result-parser-review]');
         const mcpFetchCandidateHandoffReviewRoot = document.querySelector('[data-market-intel-mcp-fetch-candidate-handoff-review]');
+        const mcpFetchCandidateQueueReviewRoot = document.querySelector('[data-market-intel-mcp-fetch-candidate-queue-review]');
         const manualSampleRoot = document.querySelector('[data-market-intel-manual-sample]');
         const sampleAcceptanceRoot = document.querySelector('[data-market-intel-sample-acceptance]');
         const sampleReviewRoot = document.querySelector('[data-market-intel-sample-review]');
@@ -1342,7 +1369,7 @@
         const liveInventoryRoot = document.querySelector('[data-market-intel-live-inventory]');
         const approvalRoot = document.querySelector('[data-market-intel-approval]');
         const deployRoot = document.querySelector('[data-market-intel-deploy]');
-        if (!root && !writerRoot && !cliRoot && !dbProbeRoot && !seedDiffRoot && !legacyBridgeRoot && !mcpReadinessRoot && !mcpPreflightRoot && !mcpActivationRoot && !mcpFetchGateRoot && !mcpCompletionRoot && !mcpActivationEvidenceRoot && !mcpRuntimeSmokeRoot && !mcpRuntimePromotionRoot && !mcpManualFetchHandoffRoot && !mcpFetchTargetReviewRoot && !mcpFetchRunPackageRoot && !mcpFetchRunReadinessRoot && !mcpFetchRunReceiptRoot && !mcpFetchResultParserReviewRoot && !mcpFetchCandidateHandoffReviewRoot && !manualSampleRoot && !sampleAcceptanceRoot && !sampleReviewRoot && !schedulerRoot && !matchReviewRoot && !opportunityRoot && !opportunityScoringRoot && !opportunityEvidenceRoot && !opportunityAlertRoot && !migrationRoot && !migrationDrillRoot && !catalogReviewRoot && !liveSmokeRoot && !liveInventoryRoot && !approvalRoot && !deployRoot) return;
+        if (!root && !writerRoot && !cliRoot && !dbProbeRoot && !seedDiffRoot && !legacyBridgeRoot && !mcpReadinessRoot && !mcpPreflightRoot && !mcpActivationRoot && !mcpFetchGateRoot && !mcpCompletionRoot && !mcpActivationEvidenceRoot && !mcpRuntimeSmokeRoot && !mcpRuntimePromotionRoot && !mcpManualFetchHandoffRoot && !mcpFetchTargetReviewRoot && !mcpFetchRunPackageRoot && !mcpFetchRunReadinessRoot && !mcpFetchRunReceiptRoot && !mcpFetchResultParserReviewRoot && !mcpFetchCandidateHandoffReviewRoot && !mcpFetchCandidateQueueReviewRoot && !manualSampleRoot && !sampleAcceptanceRoot && !sampleReviewRoot && !schedulerRoot && !matchReviewRoot && !opportunityRoot && !opportunityScoringRoot && !opportunityEvidenceRoot && !opportunityAlertRoot && !migrationRoot && !migrationDrillRoot && !catalogReviewRoot && !liveSmokeRoot && !liveInventoryRoot && !approvalRoot && !deployRoot) return;
 
         const meta = root ? root.querySelector('[data-market-intel-preview-meta]') : null;
         const body = root ? root.querySelector('[data-market-intel-preview-body]') : null;
@@ -1449,6 +1476,12 @@
         const mcpFetchCandidateHandoffReviewReview = mcpFetchCandidateHandoffReviewRoot ? mcpFetchCandidateHandoffReviewRoot.querySelector('[data-market-intel-mcp-fetch-candidate-handoff-review-review]') : null;
         const mcpFetchCandidateHandoffReviewRefresh = mcpFetchCandidateHandoffReviewRoot ? mcpFetchCandidateHandoffReviewRoot.querySelector('[data-market-intel-mcp-fetch-candidate-handoff-review-refresh]') : null;
         const mcpFetchCandidateHandoffReviewEndpoint = "{{ url_for('market_intel.market_intel_mcp_fetch_candidate_handoff_review') }}";
+        const mcpFetchCandidateQueueReviewMeta = mcpFetchCandidateQueueReviewRoot ? mcpFetchCandidateQueueReviewRoot.querySelector('[data-market-intel-mcp-fetch-candidate-queue-review-meta]') : null;
+        const mcpFetchCandidateQueueReviewBody = mcpFetchCandidateQueueReviewRoot ? mcpFetchCandidateQueueReviewRoot.querySelector('[data-market-intel-mcp-fetch-candidate-queue-review-body]') : null;
+        const mcpFetchCandidateQueueReviewInput = mcpFetchCandidateQueueReviewRoot ? mcpFetchCandidateQueueReviewRoot.querySelector('[data-market-intel-mcp-fetch-candidate-queue-review-input]') : null;
+        const mcpFetchCandidateQueueReviewReview = mcpFetchCandidateQueueReviewRoot ? mcpFetchCandidateQueueReviewRoot.querySelector('[data-market-intel-mcp-fetch-candidate-queue-review-review]') : null;
+        const mcpFetchCandidateQueueReviewRefresh = mcpFetchCandidateQueueReviewRoot ? mcpFetchCandidateQueueReviewRoot.querySelector('[data-market-intel-mcp-fetch-candidate-queue-review-refresh]') : null;
+        const mcpFetchCandidateQueueReviewEndpoint = "{{ url_for('market_intel.market_intel_mcp_fetch_candidate_queue_review') }}";
         const manualSampleMeta = manualSampleRoot ? manualSampleRoot.querySelector('[data-market-intel-manual-sample-meta]') : null;
         const manualSampleBody = manualSampleRoot ? manualSampleRoot.querySelector('[data-market-intel-manual-sample-body]') : null;
         const manualSampleRefresh = manualSampleRoot ? manualSampleRoot.querySelector('[data-market-intel-manual-sample-refresh]') : null;
@@ -3658,6 +3691,134 @@
             }
         };
 
+        const renderMcpFetchCandidateQueueReviewMeta = data => {
+            mcpFetchCandidateQueueReviewMeta.innerHTML = [
+                `mode=${data.mode || 'unknown'}`,
+                `accepted=${data.mcp_fetch_candidate_queue_review_accepted ? 'yes' : 'no'}`,
+                `gates=${data.passed_gate_count || 0}/${data.gate_count || 0}`,
+                `items=${data.review_item_count || 0}`,
+                `campaigns=${data.campaign_candidate_count || 0}`,
+                `products=${data.product_candidate_count || 0}`
+            ].map(item => `<span class="market-intel-pill">${escapeHtml(item)}</span>`).join('');
+        };
+
+        const renderMcpFetchCandidateQueueReviewBody = data => {
+            const blockers = (data.blocked_reasons || []).join(' / ');
+            const gates = data.gates || [];
+            const handoff = data.handoff_review_summary || {};
+            const review = data.candidate_queue_review_summary || {};
+            const items = review.review_items || [];
+            const steps = data.next_operator_steps || [];
+            const renderCheck = (key, label, status) => `
+                <div class="market-intel-check">
+                    <div>
+                        <strong>${escapeHtml(key)}</strong>
+                        <small>${escapeHtml(label || '')}</small>
+                    </div>
+                    <span>${escapeHtml(status)}</span>
+                </div>
+            `;
+            mcpFetchCandidateQueueReviewBody.innerHTML = `
+                <div class="market-intel-empty mb-3">此 queue review 只審核人工覆核草案；API 不建立 queue、不更新 review_state、不抓外站、不寫 DB、不掛 scheduler。${blockers ? `阻擋：${escapeHtml(blockers)}` : ''}</div>
+                <div class="market-intel-deploy-grid">
+                    <div data-market-intel-mcp-fetch-candidate-queue-review-gates>
+                        <p class="market-intel-deploy-section-title">QUEUE REVIEW GATES</p>
+                        <div class="market-intel-check-list">${
+                            gates.length
+                                ? gates.map(item => renderCheck(item.key, item.label, item.passed ? 'PASS' : 'BLOCK')).join('')
+                                : '<div class="market-intel-empty">尚未提供 queue review gates。</div>'
+                        }</div>
+                    </div>
+                    <div data-market-intel-mcp-fetch-candidate-queue-review-handoff>
+                        <p class="market-intel-deploy-section-title">HANDOFF LINK</p>
+                        <div class="market-intel-check-list">
+                            ${renderCheck('handoff_review', `${handoff.accepted ? 'accepted' : 'pending'} / candidates=${handoff.candidate_count || 0}`, handoff.accepted ? 'ACCEPTED' : 'PENDING')}
+                            ${renderCheck('handoff_boundary', 'no queue create / no DB / no scheduler', handoff.side_effects_clear ? 'CLOSED' : 'BLOCK')}
+                        </div>
+                    </div>
+                    <div data-market-intel-mcp-fetch-candidate-queue-review-items>
+                        <p class="market-intel-deploy-section-title">REVIEW ITEMS</p>
+                        <div class="market-intel-check-list">${
+                            items.slice(0, 10).map(item => renderCheck(
+                                item.candidate_key || 'candidate',
+                                `${item.candidate_type || 'candidate'} / ${item.review_state || 'missing'} / ${item.queue_write_status || 'unknown'}`,
+                                item.review_state === 'needs_review' ? 'NEEDS REVIEW' : 'BLOCK'
+                            )).join('') || '<div class="market-intel-empty">尚未提供 candidate queue review items。</div>'
+                        }</div>
+                    </div>
+                    <div data-market-intel-mcp-fetch-candidate-queue-review-policy>
+                        <p class="market-intel-deploy-section-title">REVIEW POLICY</p>
+                        <div class="market-intel-check-list">
+                            ${renderCheck('target_queue', review.target_queue || 'missing', review.target_queue === 'manual_candidate_review' ? 'MANUAL' : 'BLOCK')}
+                            ${renderCheck('review_mode', review.review_mode || 'missing', review.review_mode === 'operator_queue_review_preview' ? 'PREVIEW' : 'BLOCK')}
+                            ${renderCheck('review_state', 'all items must stay needs_review', review.all_review_states_safe ? 'SAFE' : 'BLOCK')}
+                            ${renderCheck('queue_write', 'all items must stay not_persisted', review.all_queue_write_status_preview ? 'NO WRITE' : 'BLOCK')}
+                        </div>
+                    </div>
+                    <div data-market-intel-mcp-fetch-candidate-queue-review-next>
+                        <p class="market-intel-deploy-section-title">BOUNDARY / NEXT</p>
+                        <div class="market-intel-check-list">
+                            ${renderCheck(
+                                'api_boundary',
+                                'no queue create / no review_state update / no DB / no scheduler',
+                                data.api_writes_database ? 'BLOCK' : 'CLOSED'
+                            )}
+                            ${steps.map((item, index) => renderCheck(`step_${index + 1}`, item, 'NEXT')).join('')}
+                        </div>
+                    </div>
+                </div>
+            `;
+            if (mcpFetchCandidateQueueReviewInput && !mcpFetchCandidateQueueReviewInput.value.trim() && data.sample_candidate_queue_review_package) {
+                mcpFetchCandidateQueueReviewInput.value = JSON.stringify(data.sample_candidate_queue_review_package, null, 2);
+            }
+        };
+
+        const loadMcpFetchCandidateQueueReview = async () => {
+            if (!mcpFetchCandidateQueueReviewMeta || !mcpFetchCandidateQueueReviewBody) return;
+            mcpFetchCandidateQueueReviewBody.innerHTML = '<div class="market-intel-empty">讀取 MCP Fetch Candidate Queue 審核中...</div>';
+            try {
+                const response = await fetch(mcpFetchCandidateQueueReviewEndpoint, { credentials: 'same-origin' });
+                if (!response.ok) throw new Error(`HTTP ${response.status}`);
+                const data = await response.json();
+                renderMcpFetchCandidateQueueReviewMeta(data);
+                renderMcpFetchCandidateQueueReviewBody(data);
+            } catch (error) {
+                mcpFetchCandidateQueueReviewMeta.innerHTML = '<span class="market-intel-pill">error</span>';
+                mcpFetchCandidateQueueReviewBody.innerHTML = `<div class="market-intel-empty">MCP Fetch Candidate Queue 審核讀取失敗：${escapeHtml(error.message)}</div>`;
+            }
+        };
+
+        const reviewMcpFetchCandidateQueueReview = async () => {
+            if (!mcpFetchCandidateQueueReviewMeta || !mcpFetchCandidateQueueReviewBody || !mcpFetchCandidateQueueReviewInput) return;
+            let parsed;
+            try {
+                parsed = JSON.parse(mcpFetchCandidateQueueReviewInput.value || '{}');
+            } catch (error) {
+                mcpFetchCandidateQueueReviewMeta.innerHTML = '<span class="market-intel-pill">json_error</span>';
+                mcpFetchCandidateQueueReviewBody.innerHTML = `<div class="market-intel-empty">JSON 格式錯誤：${escapeHtml(error.message)}</div>`;
+                return;
+            }
+            mcpFetchCandidateQueueReviewBody.innerHTML = '<div class="market-intel-empty">審核 MCP Fetch Candidate Queue 中...</div>';
+            try {
+                const response = await fetch(mcpFetchCandidateQueueReviewEndpoint, {
+                    method: 'POST',
+                    credentials: 'same-origin',
+                    headers: {
+                        'Content-Type': 'application/json',
+                        'X-CSRFToken': csrfToken
+                    },
+                    body: JSON.stringify({ candidate_queue_review: parsed })
+                });
+                const data = await response.json();
+                if (!response.ok && !data.mode) throw new Error(`HTTP ${response.status}`);
+                renderMcpFetchCandidateQueueReviewMeta(data);
+                renderMcpFetchCandidateQueueReviewBody(data);
+            } catch (error) {
+                mcpFetchCandidateQueueReviewMeta.innerHTML = '<span class="market-intel-pill">error</span>';
+                mcpFetchCandidateQueueReviewBody.innerHTML = `<div class="market-intel-empty">MCP Fetch Candidate Queue 審核失敗：${escapeHtml(error.message)}</div>`;
+            }
+        };
+
         const renderManualSampleMeta = data => {
             manualSampleMeta.innerHTML = [
                 `mode=${data.mode || 'unknown'}`,
@@ -13121,6 +13282,12 @@
         if (mcpFetchCandidateHandoffReviewReview) {
             mcpFetchCandidateHandoffReviewReview.addEventListener('click', reviewMcpFetchCandidateHandoffReview);
         }
+        if (mcpFetchCandidateQueueReviewRefresh) {
+            mcpFetchCandidateQueueReviewRefresh.addEventListener('click', loadMcpFetchCandidateQueueReview);
+        }
+        if (mcpFetchCandidateQueueReviewReview) {
+            mcpFetchCandidateQueueReviewReview.addEventListener('click', reviewMcpFetchCandidateQueueReview);
+        }
         if (manualSampleRefresh) {
             manualSampleRefresh.addEventListener('click', loadManualSample);
         }
@@ -13385,6 +13552,7 @@
         loadMcpFetchRunReceipt();
         loadMcpFetchResultParserReview();
         loadMcpFetchCandidateHandoffReview();
+        loadMcpFetchCandidateQueueReview();
         loadManualSample();
         loadSampleAcceptance();
         loadSampleReview();