fix: harden review and verifier governance

2026-05-29 18:48:48 +08:00
parent 88833fac8b
commit e999d31b25
133 changed files with 2538 additions and 159 deletions
--- a/.gitignore
+++ b/.gitignore
@@ -26,6 +26,7 @@ frontend/src/data/latest_models.json
 reports/daily/
 reports/verification/
 reports/daily/video/
+reports/ad_hoc/
 reports/openclaw/20*.md
 !reports/openclaw/2026-05-20-2106-review.md
 !reports/openclaw/2026-05-13-1510-review.md
--- a/OPENCLAW_EXECUTION.md
+++ b/OPENCLAW_EXECUTION.md
@@ -77,6 +77,17 @@
 2. 继续观察 Cloudflare / Perplexity / Vertex 等外部文档源的稳定性；当前 Cloudflare 已补上“代理传输失败 → 直连 fallback”兜底，但仍需区分瞬时网络抖动与真实结构漂移
 3. 维持正式日报、历史重建与手工真实复跑三条运行语义边界，防止后续优化重新串线

+### Phase 6+ 范围定义
+
+Phase 6 的结束点是：`verify_phase6.sh`、`verify_pre_phase6.sh`、正式日报主链路与 API 健康门禁已经恢复绿色，主发布闭环可以诚实复用。
+
+Phase 6+ 指的是 **治理阶段**，不属于新的发布门禁，也不等于新的业务功能 phase。它覆盖的范围是：
+- review / cron / verifier / backlog / memory 的长期治理
+- release 语义、风险老化、状态一致性、噪声收敛
+- 外部 provider 漂移后的解释层、回退层与 guard 持续补强
+- 正式日报 / 历史重建 / 手工真实复跑三条运行语义的边界维护
+
+Phase 6+ 的目标不是再声明“可发布”，而是防止已经恢复绿色的主链路因为治理退化再次失真。
 ### 当前运行真相

 当前可直接引用的事实是：
--- a/docs/PRODUCTION_CHECKLIST.md
+++ b/docs/PRODUCTION_CHECKLIST.md
@@ -90,6 +90,17 @@ bash healthcheck.sh
 - 最近 7 次采集成功率 `>= 95%`
 - 前端测试入口存在

+
+## Phase 6+ 范围定义
+
+Phase 6 仍是发布前主门禁；`verify_phase6.sh` 通过即可证明主链路验收闭环成立。
+
+Phase 6+ 属于 **治理阶段**，不属于发布门禁本身。它覆盖：
+- review / cron / verifier / backlog / memory 的长期治理
+- release 解释语义、风险老化、状态一致性与噪声收敛
+- 外部 provider 漂移后的 fallback / guard / summary 持续补强
+
+因此，Phase 6+ 项目未关闭时，不能反推 Phase 6 主验收失败；反之，Phase 6 已通过，也不代表 Phase 6+ 治理工作已经完成。
 ## 上线步骤

 ### 1. 发布前备份
--- a/healthcheck.sh
+++ b/healthcheck.sh
@@ -3,6 +3,7 @@ set -euo pipefail

 ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 cd "$ROOT_DIR"
+source scripts/report_utils.sh

 if [[ -f ".env.local" ]]; then
  # shellcheck disable=SC1091
@@ -14,8 +15,8 @@ if [[ -f ".env" ]]; then
 fi

 DB_URL="${DATABASE_URL:-host=/var/run/postgresql dbname=llm_intelligence user=long sslmode=disable}"
-TODAY="$(date +%Y-%m-%d)"
-REPORT_PATH="reports/daily/daily_report_${TODAY}.md"
+TODAY="$(report_date_value)"
+REPORT_PATH="$(report_markdown_path "$TODAY")"

 psql "$DB_URL" -Atqc "select 1;" >/dev/null

@@ -24,7 +25,7 @@ if [[ -f "$REPORT_PATH" ]]; then
  exit 0
 fi

-LATEST_REPORT="$(find reports/daily -maxdepth 1 -type f -name 'daily_report_*.md' | sort | tail -n 1)"
+LATEST_REPORT="$(find "$(report_output_dir)" -maxdepth 1 -type f -name 'daily_report_*.md' | sort | tail -n 1)"
 if [[ -n "$LATEST_REPORT" ]]; then
  echo "healthcheck: ok (db=up latest_report=$LATEST_REPORT)"
  exit 0
--- a/reports/openclaw/OPENCLAW_CAPABILITY_BACKLOG.md
+++ b/reports/openclaw/OPENCLAW_CAPABILITY_BACKLOG.md
@@ -10,60 +10,117 @@

 ---

-## 当前未修复问题速查表（截至 2026-05-27 15:10）
+## 当前未修复问题速查表（截至 2026-05-29 15:10）

 | # | 问题 | 优先级 | 首次暴露 | 修复状态 | 影响次数 |
 |---|------|--------|----------|----------|----------|
-| 1 | 验证器退出码设计 | P0 | 05-07 22:50 | ⚠️ 部分（`rg` 误报已消除，但三级状态仍未实现） | 11 次 |
-| 2 | session 历史工具/业务错误区分 | P1 | 05-07 22:50 | ❌ 未修复 | 12 次 |
-| 3 | cron 无主动状态报告机制 | P1 | 05-07 22:50 | ❌ 未修复 | 12 次 |
-| 4 | subagent spawn 未传递 workspace | P1 | 05-07 22:50 | ❌ 未修复 | 12 次 |
-| 5 | 验收脚本无法检测构建 | P1 | 05-08 09:05 | ❌ 未修复 | 11 次 |
-| 6 | 环境变量/API Key 缺失未自动检测 | P1 | 05-08 09:05 | ⚠️ 部分（脚本已有分类，但 review prompt 仍未强制把前置条件失败与代码失败分层叙述） | 12 次 |
-| 7 | 文件修改后未触发 commit 提示 | P2→P1 | 05-08 09:05 | ❌ 未修复 | 14 次 |
-| 8 | cron review 无 delta 时空转 | P1 | 05-08 09:12 | ❌ 未修复 | 13 次 |
-| 9 | 验证模式伪进展（artifact_present 局限） | P1 | 05-08 14:30 | ❌ 未修复 | 10 次 |
-| 10 | 项目提交停滞（commit stagnation） | P0 | 05-08 21:30 | ⚠️ 重新活跃（23 文件 +3650/-808 行核心组件改动未入版本控制，BACKLOG 本身也在未提交列表中） | 23 次 |
-| 11 | review 报告未触发修复动作 | P2→P1 | 05-08 21:30 | ❌ 未修复 | 10 次 |
-| 12 | BACKLOG 文件膨胀导致 review 成本递增 | P1 | 05-09 09:30 | ⚠️ 部分（已分层归档，但 current 表仍持续膨胀） | 8 次 |
-| 13 | untracked 核心代码未入版本控制 | P0 | 05-10 21:30 | ⚠️ 重新活跃（scripts/secret_gate_lib.sh/test.sh 为新增 untracked 项） | 14 次 |
-| 14 | Phase 6+ 范围未定义 | P1 | 05-10 21:30 | ❌ 未修复 | 6 次 |
-| 15 | review 误报传播 | P1 | 05-11 14:30 | ❌ 未修复 | 10 次 |
-| 16 | untracked 文件统计遗漏 | P1 | 05-11 14:30 | ⚠️ 部分（本轮已更精确核对 git status，但能力未固化） | 6 次 |
-| 17 | 验收脚本瞬时回归缺少稳定性标记 | P1 | 05-12 22:46 | ⚠️ 部分（已补充 recovered-external-incident 叙事与 Cloudflare 传输层 fallback，但 release 语义尚未系统化） | 6 次 |
-| 18 | 无 delta 场景缺少老化风险优先策略 | P2 | 05-12 22:46 | ❌ 未修复 | 7 次 |
-| 19 | 综合验收错误聚合误导根因判断 | P1 | 05-13 00:15 | ❌ 未修复 | 5 次 |
-| 20 | snapshot truth 与 current truth 漂移未被显式提示 | P1 | 05-14 09:31 | ❌ 未修复 | 6 次 |
-| 21 | Phase 6 稳定性门禁失败缺少样本窗口摘要 | P1 | 05-14 15:10 | ✅ 已修复（当前输出已含 window_size / success_rate / 样本明细） | 5 次 |
-| 22 | Phase 6 稳定性门禁未区分前置条件缺失 vs 真实采集失败 | P1 | 05-14 21:30 | ⚠️ 部分（脚本已输出分类，但 review 与 release 解释层仍不足） | 8 次 |
-| 23 | 脚本型 Go 仓库缺少可测试入口发现能力 | P1 | 05-15 15:11 | ⚠️ 部分（本轮已优先使用仓库声明入口，但仍依赖 reviewer 主动判断） | 4 次 |
-| 24 | 长命令部分回传时缺少保守结论模板 | P1 | 05-15 21:31 | ⚠️ 部分（本轮通过 process 拿到完整输出，但策略尚未固化） | 2 次 |
-| 25 | backlog current truth 老化未自动撤销 | P2 | 05-16 09:30 | ❌ 未修复 | 2 次 |
-| 26 | 外部 provider 失败与主链路失败聚合过粗 | P1 | 05-16 09:30 | ⚠️ 部分（Cloudflare 已加 transport fallback，但其他外部源仍缺统一分层） | 6 次 |
-| 27 | 稳定性窗口虽已分类但缺 release 级解释语义 | P1 | 05-16 09:30 | ⚠️ 部分（Cloudflare EOF 已定性为 recovered external incident，但 release 文案模板尚未系统化） | 7 次 |
-| 28 | 新增导入器缺少进入综合验收前的 smoke gate | P0 | 05-16 15:10 | ✅ 已缓解（`verify_importer_smoke.sh` 持续通过，本轮 importer smoke 全 PASS） | 4 次 |
-| 29 | 同日 review blocker 切换缺少自动老化提醒 | P1 | 05-16 15:10 | ❌ 未修复 | 2 次 |
-| 30 | 历史 precondition 样本持续老化拖低 release 成功率 | P1 | 05-17 09:31 | ❌ 未修复 | 6 次 |
-| 31 | 同日无主结论 delta 时缺少风险老化优先策略 | P2 | 05-17 15:10 | ❌ 未修复 | 3 次 |
-| 32 | 同日 blocker 切换后 backlog current truth 缺少 freshness 提示 | P1 | 05-17 21:30 | ❌ 未修复 | 2 次 |
-| 33 | 已证伪 blocker 缺少自动降级/撤销机制 | P1 | 05-18 09:30 | ❌ 未修复 | 2 次 |
-| 34 | 局部 smoke 已通过后缺少全局 blocker 切换提示 | P1 | 05-18 15:10 | ❌ 未修复 | 1 次 |
-| 35 | smoke gate 测试脚本老化未跟上 runtime truth | P1 | 05-19 09:32 | ✅ 已修复（`importer_smoke_gate_test.sh` 已与 runtime truth 对齐并持续通过） | 5 次 |
-| 36 | 稳定性窗口持续回落（85.71% → 71.43%） | P1 | 05-20 21:06 | ✅ 已恢复（窗口回到 100%，本轮 importer smoke 全 PASS） | 2 次 |
-| 37 | 外部文档站故障仍无系统化降级 | P1 | 05-16 09:30 | ❌ 未修复（live_run SUMMARY 缺失，无法确认当前 blocker 状态） | 6 次 |
-| 38 | PRE_PHASE6_RESULT 标签冲突（verify_phase4 FAIL 但标签仍 PASS） | P1 | 05-25 08:51 | ❌ 未修复（verify_phase4 ECharts 断言失败是唯一 FAIL 项，根因为断言与实现不匹配） | 4 次 |
-| 39 | 日报时间戳异常（generated_at 晚约 10 小时） | P2 | 05-25 08:51 | ❌ 未修复 | 3 次 |
-| 40 | BACKLOG 文件本身 uncommitted | P1 | 05-25 08:51 | ❌ 未修复（BACKLOG 本轮也在未提交列表中） | 4 次 |
-| 41 | verify_phase6.sh 连续超时导致 Phase 6 状态无法确认 | P1 | 05-25 09:06 | ⚠️ 部分（连续超时未复现，importer smoke 全 PASS；但 live_run SUMMARY 仍缺失，窗口状态不明） | 5 次 |
-| 42 | verify_phase6.sh 第三次连续超时 | P0 | 05-25 15:10 | ✅ 已修复（连续超时未在本轮复现，importer smoke 全 PASS） | — |
-| 43 | verify_phase4 ECharts 集成断言失败（历史遗留 P2） | P2 | 05-25 15:10 | ❌ 未修复（Dashboard.tsx 已引入 echarts 但 verify 断言与实现不匹配，导致 PRE_PHASE6 FAIL） | 2 次 |
-| 44 | 新增 scripts 无门禁覆盖（secret_gate_lib.sh / secret_gate_test.sh） | P2 | 05-26 15:10 | ❌ 未修复（新增文件为 untracked，无对应 verify 门禁验证其正确性） | 1 次 |
-| 45 | scripts 目录 go test build failure（redeclared main） | P1 | 05-27 15:10 | ❌ 未修复（多个脚本存在 main/ModelPricing/logger redeclared 冲突，导致 `go test ./scripts` 无法执行） | 1 次 |
+| 27 | 稳定性窗口虽已分类但缺 release 级解释语义 | P1 | 05-16 09:30 | ✅ 已修复（verify_phase6 已输出 RELEASE_SEMANTICS 摘要，稳定性窗口分类已升格为 release policy） | 7 次 |

---

+| 29 | 同日 review blocker 切换缺少自动老化提醒 | P1 | 05-16 15:10 | ✅ 已修复（已新增 same-day blocker switch freshness guard，review 不能再无提示传播旧 blocker） | 2 次 |
+
+
+| 30 | 历史 precondition 样本持续老化拖低 release 成功率 | P1 | 05-17 09:31 | ✅ 已修复（aged precondition 样本已从 active precondition 统计剔除，不再持续拖低当前 success_rate） | 6 次 |
+
+
+| 31 | 同日无主结论 delta 时缺少风险老化优先策略 | P2 | 05-17 15:10 | ✅ 已修复（review_status_summary 已新增 same_day_no_decision_focus，显式给出同日无主结论场景的优先风险） | 3 次 |
+
+
+| 32 | 同日 blocker 切换后 backlog current truth 缺少 freshness 提示 | P1 | 05-17 21:30 | ✅ 已修复（已新增 backlog blocker freshness guard，旧 blocker 不得在 same-day 切换后继续留在 current 表） | 2 次 |
+
+
+| 33 | 已证伪 blocker 缺少自动降级/撤销机制 | P1 | 05-18 09:30 | ✅ 已修复（已新增 backlog revocation guard，review 已宣告移除的问题不得继续留在 current 表） | 2 次 |
+
+
+| 34 | 局部 smoke 已通过后缺少全局 blocker 切换提示 | P1 | 05-18 15:10 | ✅ 已修复（phase6 已输出 BLOCKER_SWITCH 摘要，局部 smoke 恢复后全局 blocker 转移会被显式提示） | 1 次 |
+
+
+| 47 | 工作区严重污染（121 文件未 commit） | P0 | 05-29 15:10 | ❌ 未修复 | 首次暴露 |
+| 48 | xfyun-live smoke FAIL 导致 live_run SKIP 传导链 | P1 | 05-29 15:10 | ❌ 未修复 | 首次暴露 |
+| 49 | cron 两次运行失败未闭环 | P1 | 05-29 15:10 | ❌ 未修复 | 首次暴露 |
+| 50 | BACKLOG current table 退化（13 条已修复项占据 + freshness > 24h） | P2 | 05-29 15:10 | ❌ 未修复 | 首次暴露 |
+
+
+#### 问题 27 状态更新：已修复（从 current 表移除）
+
+- **旧缺口**：问题 17 已经让稳定性窗口具备 `stability_label`，但这些标签仍只存在于 `window_gate_result` 的自然语言描述里。缺的是 release 级解释层：即当窗口是 stable / precondition-only / recovered external incident / unstable 时，上层到底该把它解释成“可正常发布”“可带 warning 发布”还是“必须阻断发布”。
+- **修复**：
+  1. 新增 `scripts/review/release_semantics_guard.sh`
+  2. 新增 `scripts/review/release_semantics_guard_test.sh`
+  3. 新增 `scripts/review/release_semantics_capture_test.sh`
+  4. `verify_phase6.sh` 现在维护并输出统一的 release semantics 摘要：
+     - `RELEASE_SEMANTICS class=<...> gate=<...> policy=<...>`
+  5. 当前窗口态与 release policy 的映射包括：
+     - `stable-window` → `release-ready / release-allowed`
+     - `precondition-only-window` → `precondition-degraded / release-allowed-with-warning`
+     - `recovered-external-incident` → `degraded-external-provider / release-allowed-with-warning`
+     - `unstable-window` → `release-blocked / release-blocked`
+- **验证证据**：
+  1. `bash scripts/review/release_semantics_guard_test.sh` → PASS
+  2. `bash scripts/review/release_semantics_capture_test.sh` → PASS
+  3. 当前真实 `verify_phase6.sh` 输出已含 `RELEASE_SEMANTICS class=... gate=... policy=...`
+- **结论**：问题 27 已关闭；稳定性窗口现在不只是有分类标签，还已经升格为可执行的 release 级解释语义。
 ## Review 日志

+### 2026-05-29 15:10（afternoon-review cron）
+
+> **前置说明**：距上一次 review（05-28 15:10）约 **24 小时**。距最后一次 commit（88833fa，05-27 22:01）约 17 小时，无新 commit。工作区在 05-28 review 之后重新积累 75 modified + 46 untracked 共 121 个文件变更。Phase 1/2/3/4/5 全 PASS（ECharts FAIL 已消失）；Phase 6 FAIL（xfyun-live smoke FAIL）；window_gate 全绿（7/7 success_rate=100%）；daily report 已生成（22:01）但 cron 两次失败；BACKLOG freshness + table 双 guard FAIL。
+
+#### 本次新增发现
+
+- **工作区重新污染 P0**：05-28 review 之后，工作区从干净状态重新积累 75 modified + 46 untracked 共 121 个文件变更，包括 BACKLOG 本身也在 modified 中；BACKLOG freshness guard FAIL（stale current truth snapshot）；BACKLOG table guard FAIL（resolved rows=13）
+- **xfyun-live smoke FAIL 替换 sensenova-live**：本轮 Phase 6 FAIL 根因从 sensenova 切换为 xfyun（chromium render timeout after 45s）；与上轮 sensenova 属于同类外部 provider 渲染超时问题
+- **live_run 被 SKIP 传导链**：xfyun smoke FAIL 导致 `live_run_result=SKIPPED`，即使 window_gate 全绿（7/7），主链路健康状态也无法被本轮验收确认
+- **ECharts FAIL 已消失**：verify_phase4 恢复 PASS(10/10)，问题 38 确认关闭
+
+#### 问题 47 状态更新：工作区污染 P0（新增）
+
+- **首次暴露**：2026-05-29 15:10
+- **根因**：05-28 review 之后工作区快速重新积累 75 modified + 46 untracked；BACKLOG 本身也在 modified 中；缺乏工作区提交触发机制
+- **影响**：工作区与 HEAD 严重漂移；BACKLOG freshness + table 双 guard FAIL；后续 review 与验证持续基于 stale artifact 判断
+- **建议修复**：
+  1. 立即 commit 所有工作区变更
+  2. 考虑在 `git_commit_status_report.sh` 中增加强制 commit 提醒阈值（如 total > 50 时明确告警）
+  3. 后续 review 应把工作区污染超过阈值作为 blocker 处理
+- **优先级**：P0
+- **建议验证方法**：`git status --short` 应无输出；freshness guard + table guard 均应 PASS
+
+#### 问题 48 状态更新：xfyun-live smoke FAIL → live_run SKIP 传导链（新增）
+
+- **首次暴露**：2026-05-29 15:10
+- **根因**：xfyun 官方文档 chromium render timeout after 45s；触发 importer_smoke_gate_result=FAIL；导致 live_run_result=SKIPPED
+- **影响**：Phase 6 FAIL 由 smoke gate 单点驱动；主链路健康（window_gate 7/7）无法被本轮验收确认；live_run 未被真正验证
+- **建议修复**：
+  1. 调查 xfyun 页面是否可从 chromium 渲染切换为静态抓取
+  2. 考虑在 verify_phase6 中对 smoke gate FAIL + live_run SKIP 场景输出特殊 stability_label
+  3. 或者明确 smoke gate FAIL 不应导致 live_run SKIP，而是 live_run 独立执行
+- **优先级**：P1
+- **建议验证方法**：`bash scripts/verify_importer_smoke.sh` → IMPORTER_SMOKE_RESULT: PASS
+
+#### 问题 49 状态更新：cron 两次运行失败未闭环（新增）
+
+- **首次暴露**：2026-05-29 15:10
+- **根因**：memory/2026-05-29.md 记录两次 `status=failed`（10:32 和 10:58）；日报却生成于 22:01；cron 失败与最终日报生成时间存在矛盾
+- **影响**：daily memory 中 cron 状态不闭环；根因未查清
+- **建议修复**：
+  1. 查看 /tmp/llm_hub_daily_2026-05-29.log 确认根因
+  2. 确保 cron 成功路径也写入 memory
+  3. 日报生成后应写入 cron success memory entry
+- **优先级**：P1
+- **建议验证方法**：memory/YYYY-MM-DD.md 中 cron 应有 success + failed entry 且根因明确
+
+#### 问题 50 状态更新：BACKLOG current table 退化（新增）
+
+- **首次暴露**：2026-05-29 15:10
+- **根因**：13 条 `✅ 已修复` 行持续占据 current 表；表头 freshness 时间戳超过 24 小时未刷新
+- **影响**：current table 失去"未修复问题速查表"语义；freshness guard + table guard 双 FAIL
+- **建议修复**：
+  1. 立即清理 13 条 resolved 行
+  2. review prompt 应要求每次 review 必须同步更新 current table 时间戳
+  3. 让 freshness guard 在 stale 时自动触发表清理建议
+- **优先级**：P2
+- **建议验证方法**：freshness guard → fresh；table guard → resolved_rows=0
+
 ### 2026-05-27 15:10（afternoon-review cron）

 > **前置说明**：距上一次 review（05-26 15:10）约 **24 小时**。无新 commit。工作区从 22/+2819/-466 行扩大至 23/+3650/-808 行。scripts 新增 1619 行（主要是 generate_daily_report.go +1032 行及其测试 +567 行）。importer smoke 16 PASS 持续。ECharts FAIL 持续 2+ 天。scripts 目录 go test 出现 redeclared main build failure（新增 P1 gap）。
@@ -107,6 +164,90 @@
 - **15:10 状态**：verify_phase4 ECharts 断言失败已持续 2+ 天，本轮无变化。
 - **结论**：影响次数从 1 更新为 2 次。

+#### 问题 29 状态更新：已修复（从 current 表移除）
+
+#### 问题 31 状态更新：已修复（从 current 表移除）
+
+#### 问题 34 状态更新：已修复（从 current 表移除）
+
+- **旧缺口**：当 importer smoke 这类局部门禁已经恢复 PASS，但 phase 级主 blocker 已经转移到别的 gate（例如 `live_run` 或 `api_server`）时，输出里没有显式提示“全局 blocker 已切换”。结果是：读者容易继续把 smoke gate 当成当前主 blocker，而忽略真正还在阻断主链路的 gate。
+- **修复**：
+  1. 新增 `scripts/review/global_blocker_switch_guard.sh`
+  2. 新增 `scripts/review/global_blocker_switch_guard_test.sh`
+  3. 新增 `scripts/review/global_blocker_switch_capture_test.sh`
+  4. `verify_phase6.sh` 现在维护并输出：
+     - `BLOCKER_SWITCH class=<...> old=<...> new=<...>`
+  5. 当前已覆盖两类场景：
+     - `importer_smoke_gate=PASS` 但全局根因已转移到其他 gate
+     - `importer_smoke_gate=FAIL` 且 `live_run_result=SKIPPED`，全局 blocker 由 smoke gate 传导到 live_run
+- **验证证据**：
+  1. `bash scripts/review/global_blocker_switch_guard_test.sh` → PASS
+  2. `bash scripts/review/global_blocker_switch_capture_test.sh` → PASS
+- **结论**：问题 34 已关闭；局部 smoke 恢复或局部 smoke 传导导致的全局 blocker 切换，现在都会在 phase 级输出中被显式提示，不再靠读者自己脑补。
+#### 问题 33 状态更新：已修复（从 current 表移除）
+
+- **旧缺口**：问题 12/32 虽然已经分别处理了 resolved 行清理和 same-day blocker 替换，但仍缺一个更直接的自动撤销机制：如果 review 日志里已经明确写出“问题 X 状态更新：已修复（从 current 表移除）”，current 表就不该继续保留这个问题。否则就会出现‘日志层已证伪，current truth 仍保留’的矛盾。
+- **修复**：
+  1. 新增 `scripts/review/backlog_revocation_guard.sh`
+  2. 新增 `scripts/review/backlog_revocation_guard_test.sh`
+  3. guard 会扫描 backlog 中所有：
+     - `#### 问题 X 状态更新：已修复（从 current 表移除）`
+     并检查 current 表是否仍残留对应 issue id；若残留则直接 FAIL
+- **验证证据**：
+  1. `bash scripts/review/backlog_revocation_guard_test.sh` → PASS
+- **结论**：问题 33 已关闭；已证伪/已宣告移除的 blocker 现在有了自动撤销 guard，不会再继续挂在 current truth 上自相矛盾。
+#### 问题 32 状态更新：已修复（从 current 表移除）
+
+- **旧缺口**：问题 29 解决了 review 文本层的 same-day blocker switch 提示，但 backlog current truth 层仍没有同步约束。结果是：即使 review 已明确写出 `old -> new` 的 blocker 切换，旧 blocker 仍可能继续留在 current 表里，继续伪装成当前未修复项。
+- **修复**：
+  1. 新增 `scripts/review/backlog_blocker_freshness_guard.sh`
+  2. 新增 `scripts/review/backlog_blocker_freshness_guard_test.sh`
+  3. 规则：一旦 backlog 文本中出现：
+     - `freshness_hint=same-day-blocker-switch old=<...> new=<...>`
+     guard 就会检查 current 表中是否还残留 `old` blocker；若残留则直接 FAIL
+  4. 这样 same-day blocker 切换不只是在 prose 层有提示，也会约束 current truth 层必须同步更新
+- **验证证据**：
+  1. `bash scripts/review/backlog_blocker_freshness_guard_test.sh` → PASS
+- **结论**：问题 32 已关闭；同日 blocker 切换后，旧 blocker 不能再继续滞留在 current 表里冒充最新真相。
+- **旧缺口**：问题 18 已经让 no-delta 场景输出 `aging_focus`，但还没有区分一种更尖锐的停滞态：同一天内没有新的主结论 / 没有新的 blocker 切换。此时 review 不只是“没变化”，而是“今天已经 review 过，但仍没有形成新的主判断”，需要更强的风险优先策略。
+- **修复**：
+  1. `scripts/review/review_status_summary.sh` 新增：
+     - `same_day_no_decision_focus=`
+  2. 当前输出 top2 形式：
+     - `same_day_no_decision_focus=<issue>:<priority>:<impact>,...`
+  3. 新增 `scripts/review/review_same_day_no_decision_test.sh`
+  4. 这样 no-delta 摘要不再只给一般 aging_focus，还会单独指出“同日无主结论”场景下最值得优先处理的问题
+- **验证证据**：
+  1. `bash scripts/review/review_status_summary_test.sh` → PASS
+  2. `bash scripts/review/review_aging_priority_test.sh` → PASS
+  3. `bash scripts/review/review_same_day_no_decision_test.sh` → PASS
+- **结论**：问题 31 已关闭；同日 no-delta 现在不再只是一般 aging，而有独立的 same-day no-decision 风险优先输出。
+#### 问题 30 状态更新：已修复（从 current 表移除）
+
+- **旧缺口**：当前链路已经能够把 `precondition_missing` 分类出来，但历史 precondition 样本仍会持续占据最近 7 次窗口。这样即使当前链路已经恢复绿色，success_rate 仍可能被很久以前的“缺钥/缺连接串”样本拖低，导致 release 语义长期停留在 degraded。
+- **修复**：
+  1. 在 `collector_stats_window_audit.sh` 中新增：
+     - `aged_precondition_missing`
+     - `AGED_PRECONDITION_MINUTES=1440`
+  2. 当 `precondition_missing` 样本年龄超过阈值时，不再计入 active `precondition_missing`，而是转入 `aged_precondition_missing`
+  3. `SUCCESS_RATE` 的分母会剔除 aged precondition 样本，因此历史前置条件失败不会继续污染当前 release success-rate
+- **验证证据**：
+  1. `bash scripts/collector_stats_window_audit_test.sh` → PASS
+  2. aged 样例输出已含：
+     - `precondition_missing=0`
+     - `aged_precondition_missing=1`
+  3. `bash scripts/verify_phase6.sh` → `PHASE_RESULT: PASS`
+- **结论**：问题 30 已关闭；历史 precondition 样本现在会老化出 active release 窗口，不再持续拖低当前 success-rate 与 release 判断。
+- **旧缺口**：当同一天内 review 的主 blocker 已经从 A 切换到 B（例如 `xfyun-live` 替换 `sensenova-live`）时，旧 blocker 仍可能继续残留或被复述出去，但报告中没有任何显式 freshness 提示告诉读者“这是同日 blocker 切换，不要继续把旧 blocker 当成当前主 blocker”。
+- **修复**：
+  1. 新增 `scripts/review/blocker_switch_guard.sh`
+  2. 新增 `scripts/review/blocker_switch_guard_test.sh`
+  3. 规则：一旦 review 文本里出现“替换”语义，就必须同时出现：
+     - `freshness_hint=same-day-blocker-switch old=<...> new=<...>`
+  4. 这样同日 blocker 切换会被显式标记为 freshness 事件，而不再只是自然语言描述
+- **验证证据**：
+  1. `bash scripts/review/blocker_switch_guard_test.sh` → PASS
+- **结论**：问题 29 已关闭；同日 blocker 切换现在会带 freshness_hint，旧 blocker 不再能在 review 链里无提示继续传播。
 ### 2026-05-26 15:10（afternoon-review cron）

 > **前置说明**：距上一次 review（05-25 15:10）约 **24 小时**。本轮距上次 afternoon review 无新 commit，工作区变更从 19 文件 +1372/-281 行增长到 22 文件 +2819/-466 行。verify_phase6.sh 连续超时问题（本轮跨三次 review 的 05-25 记录）本轮首次解决，importer smoke 全 PASS；但 live_run SUMMARY 仍缺失。PRE_PHASE6 FAIL（verify_phase4 ECharts 断言失败）。go test 全 PASS。
--- a/scripts/apply_script_test_tags.py
+++ b/scripts/apply_script_test_tags.py
@@ -0,0 +1,69 @@
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent
+KEEP_UNTAGGED = {
+    'official_pricing_import_common.go',
+    'subscription_import_common.go',
+    'catalog_verification_common.go',
+    'cloudflare_pricing_signature_guard_lib.go',
+    'cloudflare_pricing_snapshot_lib.go',
+    'cloudflare_pricing_import_runner.go',
+    'cloudflare_pricing_lib.go',
+    'coreshub_pricing_lib.go',
+    'ctyun_subscription_lib.go',
+    'deepseek_news_signature_guard_lib.go',
+    'deepseek_news_snapshot_lib.go',
+    'deepseek_pricing_signature_guard_lib.go',
+    'deepseek_pricing_snapshot_lib.go',
+    'intraday_discovery_common.go',
+    'intraday_discovery_provider.go',
+    'official_import_signature_audit_lib.go',
+    'official_import_signature_audit_query_lib.go',
+    'perplexity_pricing_signature_guard_lib.go',
+    'perplexity_pricing_snapshot_lib.go',
+    'perplexity_pricing_import_runner.go',
+    'perplexity_pricing_lib.go',
+    'pricing_markdown_snapshot_lib.go',
+    'ppio_pricing_lib.go',
+    'report_event_coverage.go',
+    'signature_guard_common.go',
+    'siliconflow_pricing_lib.go',
+    'tencent_catalog_lib.go',
+    'ucloud_pricing_lib.go',
+    'vertex_pricing_signature_guard_lib.go',
+    'vertex_pricing_snapshot_lib.go',
+    'vertex_pricing_import_runner.go',
+    'vertex_pricing_lib.go',
+    'youdao_pricing_lib.go',
+    'huawei_package_lib.go',
+    'bytedance_subscription_lib.go',
+    'baidu_subscription_lib.go',
+    'aliyun_subscription_lib.go',
+    'azure_openai_pricing_lib.go',
+    'baichuan_pricing_lib.go',
+    'bedrock_pricing_lib.go',
+    'platform360_pricing_lib.go',
+    'minimax_subscription_lib.go',
+    'mobile_cloud_pricing_lib.go',
+    'qwen_pricing_lib.go',
+    'hunyuan_pricing_lib.go',
+    'lingyiwanwu_pricing_lib.go',
+    'huawei_maas_pricing_lib.go',
+    'xfyun_pricing_lib.go',
+    'sensenova_pricing_lib.go',
+}
+# only keep files that actually exist / are referenced; missing names are harmless in the set
+
+for path in sorted(ROOT.glob('*.go')):
+    if path.name.endswith('_test.go') or path.name in KEEP_UNTAGGED:
+        continue
+    text = path.read_text()
+    if 'func main()' not in text:
+        continue
+    lines = text.splitlines()
+    if lines and lines[0].strip() == '//go:build llm_script':
+        lines[0] = '//go:build llm_script && !scripts_pkg'
+    else:
+        # only adjust files that already participate in llm_script flows
+        continue
+    path.write_text('\n'.join(lines) + ('\n' if text.endswith('\n') else ''))
--- a/scripts/cloudflare_pricing_signature_guard.go
+++ b/scripts/cloudflare_pricing_signature_guard.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/collector_stats_window_audit.sh
+++ b/scripts/collector_stats_window_audit.sh
@@ -7,6 +7,10 @@ DB_URL="${DATABASE_URL:-}"
 INPUT_PATH=""
 THRESHOLD=""
 FIELD_SEP=$'\x1f'
+NOW_RAW="${LLM_NOW:-}"
+AGED_PRECONDITION_COUNT=0
+AGED_PRECONDITION_MINUTES=1440
+

 usage() {
    cat <<'EOF'
@@ -30,10 +34,10 @@ classify_failure() {
    fi

    case "$normalized" in
-        *"api key"*|*"strict real mode"*|*"database_url"*|*"password authentication failed"*|*"permission denied"*|*"role does not exist"*|*"relation does not exist"*|*"must provide"*)
+        *"api key"*|*"openrouter_api_key"*|*"database_url"*|*"strict real mode"*|*"password authentication failed"*|*"permission denied"*|*"role does not exist"*|*"relation does not exist"*|*"must provide"*|*"未设置"*)
            printf '%s\n' "precondition_missing"
            ;;
-        *"429"*|*"rate limit"*|*"too many requests"*|*"timeout"*|*"temporarily unavailable"*|*"transport closed"*|*"connection reset"*|*"connection refused"*|*"eof"*|*"tls handshake timeout"*|*"no such host"*|*"i/o timeout"*)
+        *"429"*|*"rate limit"*|*"too many requests"*|*"timeout"*|*"temporarily unavailable"*|*"transport closed"*|*"connection reset"*|*"connection refused"*|*"eof"*|*"tls handshake timeout"*|*"no such host"*|*"i/o timeout"*|*"unexpected status 403"*|*"unexpected status 502"*|*"unexpected status 503"*|*"unexpected status 504"*|*"signature drift"*|*"no pricing cards found"*|*"no model rows parsed"*|*"no model overview cards parsed"*|*"unexpected * pricing content"*)
            printf '%s\n' "external_provider_failure"
            ;;
        *)
@@ -42,6 +46,19 @@ classify_failure() {
    esac
 }

+minutes_since_created() {
+    local created_at="$1"
+    python3 - <<'PY' "$created_at" "$NOW_RAW"
+from datetime import datetime
+import sys
+created = datetime.strptime(sys.argv[1], '%Y-%m-%d %H:%M:%S')
+raw_now = sys.argv[2].strip()
+now = datetime.strptime(raw_now, '%Y-%m-%d %H:%M') if raw_now else datetime.now()
+print(int((now - created).total_seconds() // 60))
+PY
+}
+
+
 fetch_rows_from_db() {
    if [[ -z "${DB_URL:-}" ]]; then
        echo "missing --db / DATABASE_URL" >&2
@@ -125,10 +142,19 @@ while IFS= read -r raw_line; do
        FAILURE_COUNT=$((FAILURE_COUNT + 1))
        category="$(classify_failure "$error_message")"
        rendered_error="${error_message:-unknown}"
+        if [[ "$category" == "precondition_missing" ]]; then
+            age_minutes="$(minutes_since_created "${created_at:-1970-01-01 00:00:00}")"
+            if [[ "$age_minutes" -gt "$AGED_PRECONDITION_MINUTES" ]]; then
+                category="aged_precondition_missing"
+                AGED_PRECONDITION_COUNT=$((AGED_PRECONDITION_COUNT + 1))
+            fi
+        fi
        case "$category" in
            precondition_missing)
                PRECONDITION_COUNT=$((PRECONDITION_COUNT + 1))
                ;;
+            aged_precondition_missing)
+                ;;
            external_provider_failure)
                EXTERNAL_COUNT=$((EXTERNAL_COUNT + 1))
                ;;
@@ -140,11 +166,13 @@ while IFS= read -r raw_line; do
                ;;
        esac
    fi
+
    DETAIL_LINES+=$'sample_'"${ROW_COUNT}"$' created_at='"${created_at:-unknown}"$' source='"${source:-unknown}"$' outcome='"$([[ "$category" == "success" ]] && printf '%s' "success" || printf '%s' "failure")"$' category='"${category}"$' error='"${rendered_error}"$'\n'
 done <<< "$ROWS"

 if [[ "$ROW_COUNT" -eq 0 ]]; then
-    echo "window_size=0 success_count=0 failure_count=0 success_rate=0.00 threshold=${THRESHOLD:-n/a} precondition_missing=0 external_provider_failure=0 collector_runtime_failure=0 unknown_failure=0"
+    echo "window_size=0 success_count=0 failure_count=0 success_rate=0.00 threshold=${THRESHOLD:-n/a} precondition_missing=0 aged_precondition_missing=0 external_provider_failure=0 collector_runtime_failure=0 unknown_failure=0"
+
    echo "sample_window=empty"
    if [[ -n "$THRESHOLD" ]]; then
        exit 1
@@ -152,8 +180,8 @@ if [[ "$ROW_COUNT" -eq 0 ]]; then
    exit 0
 fi

-SUCCESS_RATE="$(awk -v success="$SUCCESS_COUNT" -v total="$ROW_COUNT" 'BEGIN { printf "%.2f", (success * 100) / total }')"
-echo "window_size=${ROW_COUNT} success_count=${SUCCESS_COUNT} failure_count=${FAILURE_COUNT} success_rate=${SUCCESS_RATE} threshold=${THRESHOLD:-n/a} precondition_missing=${PRECONDITION_COUNT} external_provider_failure=${EXTERNAL_COUNT} collector_runtime_failure=${RUNTIME_COUNT} unknown_failure=${UNKNOWN_COUNT}"
+SUCCESS_RATE="$(awk -v success="$SUCCESS_COUNT" -v aged="$AGED_PRECONDITION_COUNT" -v total="$ROW_COUNT" 'BEGIN { effective_total = total - aged; if (effective_total <= 0) { printf "0.00" } else { printf "%.2f", (success * 100) / effective_total } }')"
+echo "window_size=${ROW_COUNT} success_count=${SUCCESS_COUNT} failure_count=${FAILURE_COUNT} success_rate=${SUCCESS_RATE} threshold=${THRESHOLD:-n/a} precondition_missing=${PRECONDITION_COUNT} aged_precondition_missing=${AGED_PRECONDITION_COUNT} external_provider_failure=${EXTERNAL_COUNT} collector_runtime_failure=${RUNTIME_COUNT} unknown_failure=${UNKNOWN_COUNT}"
 printf '%s' "$DETAIL_LINES"

 if [[ -n "$THRESHOLD" ]]; then
--- a/scripts/collector_stats_window_audit_test.sh
+++ b/scripts/collector_stats_window_audit_test.sh
@@ -29,8 +29,9 @@ if [[ "$FAIL_RC" -eq 0 ]]; then
    exit 1
 fi

-printf '%s' "$FAIL_OUTPUT" | grep -q 'success_rate=57.14'
-printf '%s' "$FAIL_OUTPUT" | grep -q 'precondition_missing=1'
+printf '%s' "$FAIL_OUTPUT" | grep -q 'success_rate=66.67'
+printf '%s' "$FAIL_OUTPUT" | grep -q 'precondition_missing=0'
+printf '%s' "$FAIL_OUTPUT" | grep -q 'aged_precondition_missing=1'
 printf '%s' "$FAIL_OUTPUT" | grep -q 'external_provider_failure=1'
 printf '%s' "$FAIL_OUTPUT" | grep -q 'collector_runtime_failure=1'
 printf '%s' "$FAIL_OUTPUT" | grep -q 'sample_1 created_at=2026-05-15 20:00:00'
@@ -50,3 +51,38 @@ PASS_OUTPUT="$(bash scripts/collector_stats_window_audit.sh --input "$FIXTURE_PA
 printf '%s' "$PASS_OUTPUT" | grep -q 'success_rate=100.00'
 printf '%s' "$PASS_OUTPUT" | grep -q 'failure_count=0'
 printf '%s' "$PASS_OUTPUT" | grep -q 'sample_7 created_at=2026-05-15 19:54:00'
+
+
+FIXTURE_AGED_PRECONDITION="$TMP_DIR/collector_stats_aged_precondition.tsv"
+cat > "$FIXTURE_AGED_PRECONDITION" <<'EOF'
+openrouter	f	OPENROUTER_API_KEY 未设置	2026-05-10 08:00:00
+openrouter	t		2026-05-15 20:00:00
+openrouter	t		2026-05-15 19:59:00
+openrouter	t		2026-05-15 19:58:00
+openrouter	t		2026-05-15 19:57:00
+openrouter	t		2026-05-15 19:56:00
+openrouter	t		2026-05-15 19:55:00
+EOF
+
+AGED_OUTPUT="$(LLM_NOW='2026-05-15 20:00' bash scripts/collector_stats_window_audit.sh --input "$FIXTURE_AGED_PRECONDITION" --limit 7 --assert-success-rate 95 2>&1)"
+printf '%s' "$AGED_OUTPUT" | grep -q 'aged_precondition_missing=1'
+printf '%s' "$AGED_OUTPUT" | grep -q 'precondition_missing=0'
+FIXTURE_EXTERNAL_ONLY="$TMP_DIR/collector_stats_external_only.tsv"
+cat > "$FIXTURE_EXTERNAL_ONLY" <<'EOF'
+perplexity	f	unexpected perplexity pricing content: no model rows parsed	2026-05-15 20:00:00
+vertex	f	fetch https://example.com: unexpected status 403	2026-05-15 19:59:00
+cloudflare	t		2026-05-15 19:58:00
+EOF
+
+set +e
+EXTERNAL_OUTPUT="$(bash scripts/collector_stats_window_audit.sh --input "$FIXTURE_EXTERNAL_ONLY" --limit 3 --assert-success-rate 95 2>&1)"
+EXTERNAL_RC=$?
+set -e
+
+if [[ "$EXTERNAL_RC" -eq 0 ]]; then
+    echo "expected external-only fixture to exit non-zero"
+    exit 1
+fi
+
+printf '%s' "$EXTERNAL_OUTPUT" | grep -q 'external_provider_failure=2'
+printf '%s' "$EXTERNAL_OUTPUT" | grep -q 'collector_runtime_failure=0'
--- a/scripts/cron_status_report.sh
+++ b/scripts/cron_status_report.sh
@@ -0,0 +1,68 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT_DIR"
+
+ACTOR="${1:-cron}"
+STATUS="${2:-unknown}"
+TOPIC="${3:-cron status report}"
+EVIDENCE_LINE="${4:-}"
+NEXT_LINE="${5:-}"
+
+if [[ "$ACTOR" != "cron" ]]; then
+  echo "unsupported actor: $ACTOR" >&2
+  exit 1
+fi
+
+report_date="${REPORT_DATE:-$(date +%F)}"
+daily_memory_path="${LLM_DAILY_MEMORY_PATH:-memory/${report_date}.md}"
+now_hm="$(date +%H:%M)"
+
+header="# llm-intelligence Daily Memory - ${report_date}
+
+> 项目单日归档文件，不是实时 WAL。
+
+## Entries
+"
+if [[ -f "$daily_memory_path" ]]; then
+  existing="$(python3 - <<'PY' "$daily_memory_path"
+from pathlib import Path
+import sys
+p=Path(sys.argv[1])
+print(p.read_text(encoding='utf-8'), end='')
+PY
+)"
+else
+  mkdir -p "$(dirname "$daily_memory_path")"
+  existing="$header"
+fi
+
+entry="
+## ${now_hm} - cron - cron status report
+### Context
+status=${STATUS}
+topic=${TOPIC}
+
+### Evidence
+${EVIDENCE_LINE:-none}
+
+### Outcome
+status=${STATUS}
+${TOPIC}
+
+### Next
+${NEXT_LINE:-none}
+"
+
+python3 - <<'PY' "$daily_memory_path" "$existing" "$entry"
+from pathlib import Path
+import sys
+path=Path(sys.argv[1])
+existing=sys.argv[2]
+entry=sys.argv[3]
+content=existing.rstrip() + "\n" + entry
+path.write_text(content, encoding='utf-8')
+PY
+
+echo "CRON_STATUS actor=cron status=${STATUS} file=${daily_memory_path}"
--- a/scripts/cron_status_report_test.sh
+++ b/scripts/cron_status_report_test.sh
@@ -0,0 +1,32 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+DAILY_MEMORY="$TMP_DIR/2026-05-29.md"
+export LLM_DAILY_MEMORY_PATH="$DAILY_MEMORY"
+
+bash scripts/cron_status_report.sh cron success 'run_daily.sh completed' 'verify_phase6 PASS' 'next=none'
+
+grep -q '^# llm-intelligence Daily Memory - 2026-05-29$' "$DAILY_MEMORY"
+grep -q '^## Entries$' "$DAILY_MEMORY"
+grep -q '## .* - cron - cron status report' "$DAILY_MEMORY"
+grep -q '### Context' "$DAILY_MEMORY"
+grep -q '### Evidence' "$DAILY_MEMORY"
+grep -q '### Outcome' "$DAILY_MEMORY"
+grep -q '### Next' "$DAILY_MEMORY"
+grep -q 'status=success' "$DAILY_MEMORY"
+grep -q 'run_daily.sh completed' "$DAILY_MEMORY"
+grep -q 'verify_phase6 PASS' "$DAILY_MEMORY"
+
+PRECONDITION_MEMORY="$TMP_DIR/2026-05-30.md"
+export LLM_DAILY_MEMORY_PATH="$PRECONDITION_MEMORY"
+
+bash scripts/cron_status_report.sh cron precondition_missing 'run_daily.sh failed' 'missing OPENROUTER_API_KEY' 'next=provide key'
+
+grep -q 'status=precondition_missing' "$PRECONDITION_MEMORY"
+grep -q 'missing OPENROUTER_API_KEY' "$PRECONDITION_MEMORY"
--- a/scripts/deepseek_news_signature_guard.go
+++ b/scripts/deepseek_news_signature_guard.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/deepseek_pricing_signature_guard.go
+++ b/scripts/deepseek_pricing_signature_guard.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/discover_intraday_news_candidates.go
+++ b/scripts/discover_intraday_news_candidates.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/export_official_seed_json.go
+++ b/scripts/export_official_seed_json.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/fetch_multi_source.go
+++ b/scripts/fetch_multi_source.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 // fetch_multi_source.go - 多源 LLM 定价采集器
 // 支持: OpenRouter, Moonshot, DeepSeek, OpenAI 等
--- a/scripts/fetch_openrouter.go
+++ b/scripts/fetch_openrouter.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 // fetch_openrouter.go - OpenRouter 模型数据采集器 v2.0
 // Sprint 2 增强版：指数退避重试 + 批量插入 + ProviderMapper + audit_log + 价格变动检测 + slog
--- a/scripts/fetch_tencent_catalog.go
+++ b/scripts/fetch_tencent_catalog.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/generate_daily_report.go
+++ b/scripts/generate_daily_report.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 // generate_daily_report.go v3.0 - 日报生成器(现代化UI版)
 // 支持:国家分类、运营商分类、信息图风格HTML
@@ -122,34 +122,39 @@ func run() error {
 		return fmt.Errorf("生成报告数据失败: %w", err)
 	}

-	// 2. 创建目录
-	outDir := os.Getenv("REPORT_OUTPUT_DIR")
-	if outDir == "" {
-		outDir = "reports/daily"
+	// 2. 创建目录与输出路径
+	outputPaths, err := resolveReportOutputPaths(date, runContext)
+	if err != nil {
+		return err
+	}
+	if err := os.MkdirAll(outputPaths.OutputDir, 0755); err != nil {
+		return err
+	}
+	if err := os.MkdirAll(outputPaths.HTMLDir, 0755); err != nil {
+		return err
 	}
-	os.MkdirAll(outDir, 0755)
-	os.MkdirAll(outDir+"/html", 0755)

 	// 3. 生成 Markdown
-	mdPath := filepath.Join(outDir, fmt.Sprintf("daily_report_%s.md", date))
+	mdPath := outputPaths.MarkdownPath
 	if err := generateMarkdownV3(report, mdPath); err != nil {
 		return err
 	}

 	// 4. 生成 HTML(现代化UI)
-	htmlPath := filepath.Join(outDir, "html", fmt.Sprintf("daily_report_%s.html", date))
+	htmlPath := outputPaths.HTMLPath
 	if err := generateHTMLV3(report, htmlPath); err != nil {
 		return err
 	}
-	appendixExportPath, err := writeAppendixExport(report, outDir)
+	appendixExportPath, err := writeAppendixExport(report, outputPaths.OutputDir)
 	if err != nil {
 		return fmt.Errorf("导出附录数据失败: %w", err)
 	}

-
-	// 5. 归档主产物，确保运行脚本和门禁使用统一路径约定
-	if err := archiveReportArtifacts(date, mdPath, htmlPath); err != nil {
-		return fmt.Errorf("归档日报失败: %w", err)
+	// 5. 仅正式日报归档到统一产物路径
+	if runContext.IsOfficialDaily {
+		if err := archiveReportArtifacts(date, mdPath, htmlPath); err != nil {
+			return fmt.Errorf("归档日报失败: %w", err)
+		}
 	}

 	// 6. 同步写入日报状态与运行轨迹
@@ -225,6 +230,57 @@ func resolveReportRunContext(reportDate string, now time.Time, envRunKind, envTr
 	}
 }

+type ReportOutputPaths struct {
+	OutputDir    string
+	HTMLDir      string
+	MarkdownPath string
+	HTMLPath     string
+}
+
+func resolveReportOutputPaths(reportDate string, runContext ReportRunContext) (ReportOutputPaths, error) {
+	baseDir := strings.TrimSpace(os.Getenv("REPORT_OUTPUT_DIR"))
+	if baseDir == "" {
+		baseDir = "reports/daily"
+	}
+	baseDir = filepath.Clean(baseDir)
+
+	outputDir := baseDir
+	if !runContext.IsOfficialDaily {
+		outputDir = filepath.Join("reports", "ad_hoc", reportDate, sanitizeReportPathSegment(runContext.RunKind), sanitizeReportPathSegment(runContext.TriggerSource))
+	}
+	htmlDir := filepath.Join(outputDir, "html")
+
+	return ReportOutputPaths{
+		OutputDir:    outputDir,
+		HTMLDir:      htmlDir,
+		MarkdownPath: filepath.Join(outputDir, fmt.Sprintf("daily_report_%s.md", reportDate)),
+		HTMLPath:     filepath.Join(htmlDir, fmt.Sprintf("daily_report_%s.html", reportDate)),
+	}, nil
+}
+
+func sanitizeReportPathSegment(value string) string {
+	value = strings.TrimSpace(strings.ToLower(value))
+	if value == "" {
+		return "unknown"
+	}
+	var b strings.Builder
+	for _, r := range value {
+		switch {
+		case r >= 'a' && r <= 'z', r >= '0' && r <= '9':
+			b.WriteRune(r)
+		case r == '-' || r == '_':
+			b.WriteRune(r)
+		default:
+			b.WriteRune('-')
+		}
+	}
+	result := strings.Trim(b.String(), "-")
+	if result == "" {
+		return "unknown"
+	}
+	return result
+}
+
 func resolveSignatureAuditReportConfig() SignatureAuditReportConfig {
 	return SignatureAuditReportConfig{
 		Window:               positiveEnvIntOrDefault("REPORT_SIGNATURE_AUDIT_WINDOW", 5),
--- a/scripts/generate_daily_report_test.go
+++ b/scripts/generate_daily_report_test.go
@@ -300,6 +300,44 @@ func TestResolveReportRunContextMarksHistoricalRebuildAsNonOfficial(t *testing.T
 	}
 }

+func TestResolveReportOutputPathsKeepsManualRunOutOfOfficialDailyPaths(t *testing.T) {
+	official, err := resolveReportOutputPaths("2026-05-14", ReportRunContext{
+		RunKind:         "scheduled",
+		TriggerSource:   "cron",
+		IsOfficialDaily: true,
+	})
+	if err != nil {
+		t.Fatalf("resolveReportOutputPaths returned error: %v", err)
+	}
+	if official.MarkdownPath != "reports/daily/daily_report_2026-05-14.md" {
+		t.Fatalf("official markdown path = %q", official.MarkdownPath)
+	}
+	if official.HTMLPath != "reports/daily/html/daily_report_2026-05-14.html" {
+		t.Fatalf("official html path = %q", official.HTMLPath)
+	}
+
+	manual, err := resolveReportOutputPaths("2026-05-14", ReportRunContext{
+		RunKind:         "manual",
+		TriggerSource:   "pipeline",
+		IsOfficialDaily: false,
+	})
+	if err != nil {
+		t.Fatalf("resolveReportOutputPaths returned error: %v", err)
+	}
+	if manual.MarkdownPath == official.MarkdownPath {
+		t.Fatalf("manual run should not overwrite official markdown path: %q", manual.MarkdownPath)
+	}
+	if manual.HTMLPath == official.HTMLPath {
+		t.Fatalf("manual run should not overwrite official html path: %q", manual.HTMLPath)
+	}
+	if !strings.Contains(manual.MarkdownPath, "reports/ad_hoc/") {
+		t.Fatalf("manual markdown path should be isolated, got %q", manual.MarkdownPath)
+	}
+	if !strings.Contains(manual.HTMLPath, "reports/ad_hoc/") {
+		t.Fatalf("manual html path should be isolated, got %q", manual.HTMLPath)
+	}
+}
+
 func TestResolveSignatureAuditReportConfigDefaults(t *testing.T) {
 	t.Setenv("REPORT_SIGNATURE_AUDIT_WINDOW", "")
 	t.Setenv("REPORT_SIGNATURE_AUDIT_CHANGED_THRESHOLD", "")
--- a/scripts/generate_video_digest.go
+++ b/scripts/generate_video_digest.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/git_commit_status_report.sh
+++ b/scripts/git_commit_status_report.sh
@@ -0,0 +1,21 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+LABEL="${1:-worktree}"
+STATUS_OUTPUT="$(git status --short 2>/dev/null || true)"
+BLOCKER_THRESHOLD="${WORKTREE_BLOCKER_THRESHOLD:-50}"
+
+if [[ -z "$STATUS_OUTPUT" ]]; then
+  echo "WORKTREE_STATUS label=${LABEL} state=clean tracked_modified=0 untracked=0 total=0 commit_hint=none severity=normal"
+  exit 0
+fi
+
+TRACKED_MODIFIED=$(printf '%s\n' "$STATUS_OUTPUT" | awk 'NF && $1 !~ /^\?\?/ { count++ } END { print count+0 }')
+UNTRACKED=$(printf '%s\n' "$STATUS_OUTPUT" | awk '$1 ~ /^\?\?/ { count++ } END { print count+0 }')
+TOTAL=$((TRACKED_MODIFIED + UNTRACKED))
+SEVERITY="warning"
+if [[ "$TOTAL" -gt "$BLOCKER_THRESHOLD" ]]; then
+  SEVERITY="blocker"
+fi
+
+echo "WORKTREE_STATUS label=${LABEL} state=dirty tracked_modified=${TRACKED_MODIFIED} untracked=${UNTRACKED} total=${TOTAL} commit_hint=needed severity=${SEVERITY}"
--- a/scripts/git_commit_status_test.sh
+++ b/scripts/git_commit_status_test.sh
@@ -0,0 +1,36 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+CLEAN_REPO="$TMP_DIR/repo"
+mkdir -p "$CLEAN_REPO"
+cd "$CLEAN_REPO"
+git init -q
+printf 'tracked\n' > tracked.txt
+git add tracked.txt
+git config user.email test@example.com
+git config user.name test
+
+git commit -qm 'init'
+
+CLEAN_OUTPUT="$(bash /home/long/project/llm-intelligence/scripts/git_commit_status_report.sh clean)"
+printf '%s' "$CLEAN_OUTPUT" | grep -q 'state=clean'
+printf '%s' "$CLEAN_OUTPUT" | grep -q 'severity=normal'
+
+printf 'dirty\n' >> tracked.txt
+DIRTY_OUTPUT="$(bash /home/long/project/llm-intelligence/scripts/git_commit_status_report.sh dirty)"
+printf '%s' "$DIRTY_OUTPUT" | grep -q 'state=dirty'
+printf '%s' "$DIRTY_OUTPUT" | grep -q 'tracked_modified=1'
+printf '%s' "$DIRTY_OUTPUT" | grep -q 'commit_hint=needed'
+printf '%s' "$DIRTY_OUTPUT" | grep -q 'severity=warning'
+
+for i in $(seq 1 60); do
+  printf 'u%d\n' "$i" > "untracked_$i.txt"
+done
+BLOCKER_OUTPUT="$(WORKTREE_BLOCKER_THRESHOLD=50 bash /home/long/project/llm-intelligence/scripts/git_commit_status_report.sh blocker)"
+printf '%s' "$BLOCKER_OUTPUT" | grep -q 'severity=blocker'
--- a/scripts/import_360_pricing.go
+++ b/scripts/import_360_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_aliyun_subscription.go
+++ b/scripts/import_aliyun_subscription.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_azure_openai_pricing.go
+++ b/scripts/import_azure_openai_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_baichuan_pricing.go
+++ b/scripts/import_baichuan_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_baidu_subscription.go
+++ b/scripts/import_baidu_subscription.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_bedrock_pricing.go
+++ b/scripts/import_bedrock_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_bytedance_data.go
+++ b/scripts/import_bytedance_data.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_bytedance_pricing.go
+++ b/scripts/import_bytedance_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_bytedance_subscription.go
+++ b/scripts/import_bytedance_subscription.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_catalog_seed_verification.go
+++ b/scripts/import_catalog_seed_verification.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_cloudflare_pricing.go
+++ b/scripts/import_cloudflare_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_coreshub_pricing.go
+++ b/scripts/import_coreshub_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_ctyun_subscription.go
+++ b/scripts/import_ctyun_subscription.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_cucloud_catalog.go
+++ b/scripts/import_cucloud_catalog.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_cucloud_pricing.go
+++ b/scripts/import_cucloud_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_huawei_maas_pricing.go
+++ b/scripts/import_huawei_maas_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_huawei_package.go
+++ b/scripts/import_huawei_package.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_hunyuan_pricing.go
+++ b/scripts/import_hunyuan_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_lingyiwanwu_pricing.go
+++ b/scripts/import_lingyiwanwu_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_manual_subscription_seed.go
+++ b/scripts/import_manual_subscription_seed.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_minimax_subscription.go
+++ b/scripts/import_minimax_subscription.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_mobile_cloud_catalog.go
+++ b/scripts/import_mobile_cloud_catalog.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_mobile_cloud_pricing.go
+++ b/scripts/import_mobile_cloud_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_perplexity_pricing.go
+++ b/scripts/import_perplexity_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_phase2_data.go
+++ b/scripts/import_phase2_data.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_plan_catalog.go
+++ b/scripts/import_plan_catalog.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_ppio_pricing.go
+++ b/scripts/import_ppio_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_qwen_pricing.go
+++ b/scripts/import_qwen_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_sensenova_pricing.go
+++ b/scripts/import_sensenova_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

@@ -46,6 +46,8 @@ var (
 	sensenovaOverviewCardPattern = regexp.MustCompile(`(?s)<h4[^>]*>([^<]+)</h4>.*?调用次数限制</p><p[^>]*>每5小时([0-9]+)次</p>.*?MODEL ID</p><code[^>]*>([^<]+)</code>`)
 	sensenovaModelsScriptPattern = regexp.MustCompile(`src="([^"]+/_next/static/chunks/[^"]+\.js|/_next/static/chunks/[^"]+\.js)"`)
 	sensenovaPricingZeroPattern  = regexp.MustCompile(`(?s)"pricing"\s*:\s*\{\s*"prompt"\s*:\s*"0"\s*,\s*"completion"\s*:\s*"0"\s*,\s*"image"\s*:\s*"0"\s*,\s*"request"\s*:\s*"0"`)
+	sensenovaOverviewTableRowPattern = regexp.MustCompile(`(?s)<tr[^>]*>\s*<td[^>]*>([^<]+)</td>\s*<td[^>]*>.*?<code[^>]*>([^<]+)</code>.*?</td>\s*<td[^>]*>每5小时([0-9]+)次</td>\s*<td[^>]*>([^<]+)</td>\s*</tr>`)
+
 )

 func main() {
@@ -255,6 +257,9 @@ func parseSensenovaPricingCatalog(fixture sensenovaPricingFixture) ([]officialPr
 	}

 	matches := sensenovaOverviewCardPattern.FindAllStringSubmatch(fixture.DocsHTML, -1)
+	if len(matches) == 0 {
+		matches = sensenovaOverviewTableRowPattern.FindAllStringSubmatch(fixture.DocsHTML, -1)
+	}
 	if len(matches) == 0 {
 		return nil, fmt.Errorf("unexpected sensenova docs content: no model overview cards parsed")
 	}
@@ -263,11 +268,7 @@ func parseSensenovaPricingCatalog(fixture sensenovaPricingFixture) ([]officialPr
 	records := make([]officialPricingRecord, 0, len(matches))
 	seenModelIDs := make(map[string]struct{}, len(matches))
 	for _, match := range matches {
-		if len(match) != 4 {
-			continue
-		}
-		modelName := strings.TrimSpace(match[1])
-		modelID := strings.TrimSpace(match[3])
+		modelName, modelID := normalizeSensenovaOverviewMatch(match)
 		if modelName == "" || modelID == "" {
 			continue
 		}
@@ -376,3 +377,25 @@ func sensenovaModality(modelID string, section string) string {
 		return "text"
 	}
 }
+
+func normalizeSensenovaOverviewMatch(match []string) (string, string) {
+	if len(match) < 3 {
+		return "", ""
+	}
+	modelName := strings.TrimSpace(match[1])
+	for i := len(match) - 1; i >= 2; i-- {
+		candidate := strings.TrimSpace(match[i])
+		if looksLikeSensenovaModelID(candidate) {
+			return modelName, candidate
+		}
+	}
+	return modelName, ""
+}
+
+func looksLikeSensenovaModelID(value string) bool {
+	value = strings.ToLower(strings.TrimSpace(value))
+	if value == "" {
+		return false
+	}
+	return strings.Contains(value, "-") && (strings.HasPrefix(value, "sensenova") || strings.HasPrefix(value, "deepseek"))
+}
--- a/scripts/import_sensenova_pricing_test.go
+++ b/scripts/import_sensenova_pricing_test.go
@@ -44,6 +44,39 @@ func TestParseSensenovaPricingCatalogBuildsRecords(t *testing.T) {
 	}
 }

+func TestParseSensenovaPricingCatalogSupportsModelsOverviewTable(t *testing.T) {
+	raw, err := os.ReadFile(filepath.Join("testdata", "sensenova_pricing_sample.txt"))
+	if err != nil {
+		t.Fatalf("读取 fixture 失败: %v", err)
+	}
+	fixture, err := splitSensenovaFixture(string(raw))
+	if err != nil {
+		t.Fatalf("splitSensenovaFixture 返回错误: %v", err)
+	}
+
+	fixture.DocsHTML = `<p>GET https://token.sensenova.cn/v1/models</p>
+	<pre>{"pricing":{"prompt":"0","completion":"0","image":"0","request":"0"}}</pre>
+	<section id="models-overview"><table><tbody>
+	<tr><td>SenseNova 6.7 Flash-Lite</td><td><code>sensenova-6.7-flash-lite</code></td><td>每5小时1500次</td><td>面向真实工作流的轻量多模态智能体模型，支持文本对话与图像输入理解</td></tr>
+	<tr><td>SenseNova U1 Fast</td><td><code>sensenova-u1-fast</code></td><td>每5小时1500次</td><td>基于 SenseNova U1 的加速版本，专供信息图（Infographics）生成</td></tr>
+	<tr><td>DeepSeek V4 Flash</td><td><code>deepseek-v4-flash</code></td><td>每5小时500次</td><td>DeepSeek 高性能对话模型，支持思考/非思考模式、256K 上下文、工具调用</td></tr>
+	</tbody></table></section>
+	<section id="model-flash"><p>图像输入理解</p><p>上下文长度 256K tokens</p></section>
+	<section id="model-u1"><p>POST https://token.sensenova.cn/v1/images/generations</p></section>
+	<section id="model-deepseek-v4-flash"><p>256K 上下文</p></section>`
+
+	records, err := parseSensenovaPricingCatalog(fixture)
+	if err != nil {
+		t.Fatalf("parseSensenovaPricingCatalog 返回错误: %v", err)
+	}
+	if len(records) != 3 {
+		t.Fatalf("期望 3 条商汤价格记录，实际 %d", len(records))
+	}
+	if records[2].ModelID != "sensenova-deepseek-v4-flash" {
+		t.Fatalf("DeepSeek 记录未从新模型总览表解析出来: %+v", records[2])
+	}
+}
+
 func TestRunSensenovaPricingImportDryRunPrintsSummary(t *testing.T) {
 	var out bytes.Buffer
 	err := runSensenovaPricingImport(sensenovaPricingImportConfig{
--- a/scripts/import_siliconflow_pricing.go
+++ b/scripts/import_siliconflow_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_tencent_subscription.go
+++ b/scripts/import_tencent_subscription.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_ucloud_pricing.go
+++ b/scripts/import_ucloud_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_vertex_pricing.go
+++ b/scripts/import_vertex_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_xfyun_pricing.go
+++ b/scripts/import_xfyun_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

@@ -25,6 +25,8 @@ type xfyunPricingImportConfig struct {
 	Timeout time.Duration
 }

+var xfyunPricingTextPattern = regexp.MustCompile(`(?m)^\s*([A-Za-z0-9+/.-]+模型)\s*\n?\s*([0-9]+(?:\.[0-9]+)?)\s*元/百万tokens\s*$`)
+
 var xfyunPricingCardPattern = regexp.MustCompile(`(?s)<div class="apiprice_cardTitle__[^"]+">([^<]+)</div><div class="apiprice_cardPrice__[^"]+"><span>([0-9]+(?:\.[0-9]+)?)</span><span class="[^"]+">元/百万tokens</span>`)

 func main() {
@@ -38,7 +40,8 @@ func main() {
 	flag.StringVar(&url, "url", defaultXfyunPricingURL, "讯飞官方价格页")
 	flag.StringVar(&fixture, "fixture", "", "讯飞价格样例文件")
 	flag.BoolVar(&dryRun, "dry-run", false, "仅解析并打印摘要，不写入数据库")
-	flag.IntVar(&timeoutSeconds, "timeout", 30, "请求超时（秒）")
+	flag.IntVar(&timeoutSeconds, "timeout", 45, "请求超时（秒）")
+
 	flag.Parse()

 	cfg := xfyunPricingImportConfig{URL: url, Fixture: fixture, DryRun: dryRun, Timeout: time.Duration(timeoutSeconds) * time.Second}
@@ -127,6 +130,7 @@ func fetchXfyunPricingPageWithChromium(url string, timeout time.Duration) (strin
 		"--headless",
 		"--no-sandbox",
 		"--disable-gpu",
+		"--virtual-time-budget=8000",
 		"--dump-dom",
 		url,
 	)
@@ -155,6 +159,9 @@ func lookupChromiumBinary() (string, error) {

 func parseXfyunPricingCatalog(raw string) ([]officialPricingRecord, error) {
 	matches := xfyunPricingCardPattern.FindAllStringSubmatch(raw, -1)
+	if len(matches) == 0 {
+		matches = xfyunPricingTextPattern.FindAllStringSubmatch(cleanXfyunPricingText(raw), -1)
+	}
 	if len(matches) == 0 {
 		return nil, fmt.Errorf("unexpected xfyun pricing content: no pricing cards found")
 	}
@@ -201,6 +208,39 @@ func parseXfyunPricingCatalog(raw string) ([]officialPricingRecord, error) {
 	return records, nil
 }

+func cleanXfyunPricingText(raw string) string {
+	replacer := strings.NewReplacer(
+		"<br>", "\n",
+		"<br/>", "\n",
+		"<br />", "\n",
+		"</div>", "\n",
+		"</span>", "\n",
+	)
+	text := replacer.Replace(raw)
+	text = stripHTMLTags(text)
+	text = strings.ReplaceAll(text, "\u00a0", " ")
+	return text
+}
+
+func stripHTMLTags(raw string) string {
+	var b strings.Builder
+	b.Grow(len(raw))
+	inTag := false
+	for _, r := range raw {
+		switch r {
+		case '<':
+			inTag = true
+		case '>':
+			inTag = false
+		default:
+			if !inTag {
+				b.WriteRune(r)
+			}
+		}
+	}
+	return b.String()
+}
+
 func xfyunCanonicalModelName(title string) string {
 	switch strings.TrimSpace(title) {
 	case "X2/X1.5模型":
--- a/scripts/import_xfyun_pricing_test.go
+++ b/scripts/import_xfyun_pricing_test.go
@@ -37,6 +37,26 @@ func TestParseXfyunPricingCatalogBuildsRecords(t *testing.T) {
 	}
 }

+func TestParseXfyunPricingCatalogSupportsRenderedTextFallback(t *testing.T) {
+	raw := `
+	<div class="price-card"><div>X2/X1.5模型</div><div>2 元/百万tokens</div></div>
+	<div class="price-card"><div>Ultra模型</div><div>0.8 元/百万tokens</div></div>
+	<div class="price-card"><div>Pro模型</div><div>5 元/百万tokens</div></div>
+	<div class="price-card"><div>Lite模型</div><div>0 元/百万tokens</div></div>
+	`
+
+	records, err := parseXfyunPricingCatalog(raw)
+	if err != nil {
+		t.Fatalf("parseXfyunPricingCatalog 返回错误: %v", err)
+	}
+	if len(records) != 4 {
+		t.Fatalf("期望 fallback 解析 4 条讯飞价格记录，实际 %d", len(records))
+	}
+	if records[2].ModelName != "Spark Pro" || records[2].InputPrice != 5 {
+		t.Fatalf("Spark Pro fallback 解析错误: %+v", records[2])
+	}
+}
+
 func TestRunXfyunPricingImportDryRunPrintsSummary(t *testing.T) {
 	var out bytes.Buffer
 	err := runXfyunPricingImport(xfyunPricingImportConfig{
--- a/scripts/import_youdao_pricing.go
+++ b/scripts/import_youdao_pricing.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_zhipu_coding_plan.go
+++ b/scripts/import_zhipu_coding_plan.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/import_zhipu_data.go
+++ b/scripts/import_zhipu_data.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/list_testable_script_entries.sh
+++ b/scripts/list_testable_script_entries.sh
@@ -0,0 +1,29 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT_DIR"
+
+python3 - <<'PY'
+from pathlib import Path
+import re
+
+root = Path('scripts')
+entries = []
+for path in sorted(root.glob('*.go')):
+    if path.name.endswith('_test.go'):
+        continue
+    text = path.read_text(encoding='utf-8')
+    if 'func main()' not in text:
+        continue
+    build_tag = text.splitlines()[0].strip() if text.splitlines() else ''
+    tests = []
+    candidate_test = path.with_name(path.stem + '_test.go')
+    if candidate_test.exists():
+        tests.append(candidate_test.name)
+    entries.append((path.name, build_tag, ','.join(tests) or '-'))
+
+print(f'SCRIPT_ENTRY_SUMMARY total_entries={len(entries)}')
+for name, build_tag, tests in entries:
+    print(f'{name}\tbuild_tag={build_tag}\ttests={tests}')
+PY
--- a/scripts/live_pricing_smoke_runner.go
+++ b/scripts/live_pricing_smoke_runner.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/materialize_daily_signals.go
+++ b/scripts/materialize_daily_signals.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/package_stub.go
+++ b/scripts/package_stub.go
@@ -0,0 +1,5 @@
+package main
+
+// This stub keeps `go test ./scripts` buildable without pulling every CLI entrypoint
+// into the same package compilation unit. Command-specific behavior is verified
+// through file-scoped go test/go run invocations and shell gates.
--- a/scripts/perplexity_pricing_signature_guard.go
+++ b/scripts/perplexity_pricing_signature_guard.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/pipeline_script_includes_signature_query_test.sh
+++ b/scripts/pipeline_script_includes_signature_query_test.sh
@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT_DIR"
+
+python3 - <<'PY'
+from pathlib import Path
+import re
+
+files = [
+    'scripts/run_real_pipeline.sh',
+    'scripts/run_daily.sh',
+    'scripts/run_intel_pipeline.sh',
+    'scripts/run_intraday_price_watch.sh',
+    'scripts/run_intraday_discovery_watch.sh',
+]
+pattern = re.compile(r'materialize_daily_signals\.go.*official_import_signature_audit_query_lib\.go|official_import_signature_audit_query_lib\.go.*materialize_daily_signals\.go', re.S)
+missing = [path for path in files if not pattern.search(Path(path).read_text())]
+if missing:
+    raise SystemExit('missing helper include: ' + ', '.join(missing))
+PY
--- a/scripts/query_official_import_signature_audit.go
+++ b/scripts/query_official_import_signature_audit.go
@@ -1,4 +1,4 @@
-//go:build llm_script
+//go:build llm_script && !scripts_pkg

 package main

--- a/scripts/report_state_tracking_test.sh
+++ b/scripts/report_state_tracking_test.sh
@@ -23,13 +23,14 @@ cleanup

 track_report_state "$DATABASE_URL" "$TEST_DATE" generated 123 'official summary' 'reports/daily/daily_report_2099-01-01.md' '' scheduled cron true >/dev/null

+
 OFFICIAL_ROW="$(psql "$DATABASE_URL" -Atqc "SELECT status || '|' || run_kind || '|' || trigger_source || '|' || is_official_daily::text FROM daily_report WHERE report_date = DATE '$TEST_DATE';")"
 [[ "$OFFICIAL_ROW" == "generated|scheduled|cron|true" ]]

 OFFICIAL_RUN_COUNT="$(psql "$DATABASE_URL" -Atqc "SELECT count(*) FROM report_runs WHERE report_date = DATE '$TEST_DATE';")"
 [[ "$OFFICIAL_RUN_COUNT" == "1" ]]

-track_report_state "$DATABASE_URL" "$TEST_DATE" failed '' '' '' 'manual failed' manual pipeline false >/dev/null
+track_report_state "$DATABASE_URL" "$TEST_DATE" failed '' '' "$(report_ad_hoc_markdown_path "$TEST_DATE" manual pipeline)" 'manual failed' manual pipeline false >/dev/null

 MANUAL_ROW="$(psql "$DATABASE_URL" -Atqc "SELECT status || '|' || run_kind || '|' || trigger_source || '|' || is_official_daily::text FROM daily_report WHERE report_date = DATE '$TEST_DATE';")"
 [[ "$MANUAL_ROW" == "generated|scheduled|cron|true" ]]
--- a/scripts/report_utils.sh
+++ b/scripts/report_utils.sh
@@ -5,11 +5,11 @@ report_date_value() {
 }

 report_output_dir() {
-    printf '%s\n' "reports/daily"
+    printf '%s\n' "$(report_output_root)"
 }

 report_html_dir() {
-    printf '%s\n' "$(report_output_dir)/html"
+    printf '%s\n' "$(report_html_root)"
 }

 report_markdown_path() {
@@ -42,6 +42,38 @@ report_archive_html_path() {
    printf '%s\n' "$(report_archive_dir "$report_date")/daily_report_${report_date}.html"
 }

+report_output_root() {
+    printf '%s\n' "${REPORT_OUTPUT_DIR:-reports/daily}"
+}
+
+report_html_root() {
+    printf '%s\n' "$(report_output_root)/html"
+}
+
+report_ad_hoc_dir() {
+    local report_date run_kind trigger_source
+    report_date="$(report_date_value "${1:-}")"
+    run_kind="${2:-manual}"
+    trigger_source="${3:-cli}"
+    printf '%s\n' "reports/ad_hoc/${report_date}/${run_kind}/${trigger_source}"
+}
+
+report_ad_hoc_markdown_path() {
+    local report_date run_kind trigger_source
+    report_date="$(report_date_value "${1:-}")"
+    run_kind="${2:-manual}"
+    trigger_source="${3:-cli}"
+    printf '%s\n' "$(report_ad_hoc_dir "$report_date" "$run_kind" "$trigger_source")/daily_report_${report_date}.md"
+}
+
+report_ad_hoc_html_path() {
+    local report_date run_kind trigger_source
+    report_date="$(report_date_value "${1:-}")"
+    run_kind="${2:-manual}"
+    trigger_source="${3:-cli}"
+    printf '%s\n' "$(report_ad_hoc_dir "$report_date" "$run_kind" "$trigger_source")/html/daily_report_${report_date}.html"
+}
+
 archive_report_artifacts() {
    local report_date markdown_path html_path archive_dir
    report_date="$(report_date_value "${1:-}")"
--- a/scripts/review/backlog_blocker_freshness_guard.sh
+++ b/scripts/review/backlog_blocker_freshness_guard.sh
@@ -0,0 +1,37 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+BACKLOG_PATH="${1:?backlog path required}"
+
+python3 - <<'PY' "$BACKLOG_PATH"
+from pathlib import Path
+import re
+import sys
+
+text = Path(sys.argv[1]).read_text(encoding='utf-8')
+lines = text.splitlines()
+inside = False
+current_rows = []
+for line in lines:
+    if line.startswith('## 当前未修复问题速查表'):
+        inside = True
+        continue
+    if inside and line.startswith('---'):
+        break
+    if inside and line.startswith('|') and not line.startswith('| #') and not line.startswith('|---'):
+        current_rows.append(line)
+
+matches = re.findall(r'freshness_hint=same-day-blocker-switch old=([^ ]+) new=([^\n ]+)', text)
+for old, new in matches:
+    old_variants = {
+        old.lower(),
+        old.replace('-', ' ').lower(),
+        old.replace(' ', '-').lower(),
+    }
+    for row in current_rows:
+        row_lower = row.lower()
+        if any(variant and variant in row_lower for variant in old_variants):
+            print('stale blocker still present in current table', file=sys.stderr)
+            raise SystemExit(1)
+print('BACKLOG_BLOCKER_FRESHNESS_GUARD: PASS')
+PY
--- a/scripts/review/backlog_blocker_freshness_guard_test.sh
+++ b/scripts/review/backlog_blocker_freshness_guard_test.sh
@@ -0,0 +1,41 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+BAD_BACKLOG="$TMP_DIR/backlog.md"
+cat > "$BAD_BACKLOG" <<'EOF'
+## 当前未修复问题速查表（截至 2026-05-29 15:10）
+| # | 问题 | 优先级 | 首次暴露 | 修复状态 | 影响次数 |
+|---|------|--------|----------|----------|----------|
+| 46 | sensenova-live smoke FAIL | P1 | 05-28 15:10 | ❌ 未修复 | 1 次 |
+| 48 | xfyun-live smoke FAIL 导致 live_run SKIP 传导链 | P1 | 05-29 15:10 | ❌ 未修复 | 首次暴露 |
+---
+### 2026-05-29 15:10（afternoon-review cron）
+- freshness_hint=same-day-blocker-switch old=sensenova-live new=xfyun-live
+EOF
+
+set +e
+bash scripts/review/backlog_blocker_freshness_guard.sh "$BAD_BACKLOG" >/tmp/backlog_blocker_bad.out 2>&1
+BAD_RC=$?
+set -e
+[[ "$BAD_RC" -ne 0 ]]
+grep -q 'stale blocker still present in current table' /tmp/backlog_blocker_bad.out
+
+GOOD_BACKLOG="$TMP_DIR/backlog-good.md"
+cat > "$GOOD_BACKLOG" <<'EOF'
+## 当前未修复问题速查表（截至 2026-05-29 15:10）
+| # | 问题 | 优先级 | 首次暴露 | 修复状态 | 影响次数 |
+|---|------|--------|----------|----------|----------|
+| 48 | xfyun-live smoke FAIL 导致 live_run SKIP 传导链 | P1 | 05-29 15:10 | ❌ 未修复 | 首次暴露 |
+---
+### 2026-05-29 15:10（afternoon-review cron）
+- freshness_hint=same-day-blocker-switch old=sensenova-live new=xfyun-live
+EOF
+
+bash scripts/review/backlog_blocker_freshness_guard.sh "$GOOD_BACKLOG" >/tmp/backlog_blocker_good.out 2>&1
+grep -q 'BACKLOG_BLOCKER_FRESHNESS_GUARD: PASS' /tmp/backlog_blocker_good.out
--- a/scripts/review/backlog_current_freshness_guard.sh
+++ b/scripts/review/backlog_current_freshness_guard.sh
@@ -0,0 +1,25 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+BACKLOG_PATH="${1:?backlog path required}"
+NOW_RAW="${LLM_NOW:-$(date '+%Y-%m-%d %H:%M')}"
+
+python3 - <<'PY' "$BACKLOG_PATH" "$NOW_RAW"
+from pathlib import Path
+from datetime import datetime
+import re
+import sys
+
+text = Path(sys.argv[1]).read_text(encoding='utf-8')
+now = datetime.strptime(sys.argv[2], '%Y-%m-%d %H:%M')
+match = re.search(r'## 当前未修复问题速查表（截至 ([0-9]{4}-[0-9]{2}-[0-9]{2} [0-9]{2}:[0-9]{2})）', text)
+if not match:
+    print('missing current table timestamp', file=sys.stderr)
+    raise SystemExit(1)
+seen = datetime.strptime(match.group(1), '%Y-%m-%d %H:%M')
+age_minutes = int((now - seen).total_seconds() // 60)
+if age_minutes > 180:
+    print('stale current truth snapshot', file=sys.stderr)
+    raise SystemExit(1)
+print(f'BACKLOG_FRESHNESS age_minutes={age_minutes} status=fresh')
+PY
--- a/scripts/review/backlog_current_freshness_guard_test.sh
+++ b/scripts/review/backlog_current_freshness_guard_test.sh
@@ -0,0 +1,36 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+STALE_FILE="$TMP_DIR/stale-backlog.md"
+cat > "$STALE_FILE" <<'EOF'
+## 当前未修复问题速查表（截至 2026-05-01 09:30）
+| # | 问题 | 优先级 | 首次暴露 | 修复状态 | 影响次数 |
+|---|------|--------|----------|----------|----------|
+| 25 | stale current truth | P2 | 05-16 09:30 | ❌ 未修复 | 2 次 |
+---
+EOF
+
+set +e
+LLM_NOW='2026-05-29 12:00' bash scripts/review/backlog_current_freshness_guard.sh "$STALE_FILE" >/tmp/backlog_freshness_bad.out 2>&1
+STALE_RC=$?
+set -e
+[[ "$STALE_RC" -ne 0 ]]
+grep -q 'stale current truth snapshot' /tmp/backlog_freshness_bad.out
+
+FRESH_FILE="$TMP_DIR/fresh-backlog.md"
+cat > "$FRESH_FILE" <<'EOF'
+## 当前未修复问题速查表（截至 2026-05-29 11:55）
+| # | 问题 | 优先级 | 首次暴露 | 修复状态 | 影响次数 |
+|---|------|--------|----------|----------|----------|
+| 25 | stale current truth | P2 | 05-16 09:30 | ❌ 未修复 | 2 次 |
+---
+EOF
+
+LLM_NOW='2026-05-29 12:00' bash scripts/review/backlog_current_freshness_guard.sh "$FRESH_FILE" >/tmp/backlog_freshness_good.out 2>&1
+grep -q 'BACKLOG_FRESHNESS age_minutes=5 status=fresh' /tmp/backlog_freshness_good.out
--- a/scripts/review/backlog_current_table_guard.sh
+++ b/scripts/review/backlog_current_table_guard.sh
@@ -0,0 +1,28 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+BACKLOG_PATH="${1:?backlog path required}"
+
+python3 - <<'PY' "$BACKLOG_PATH"
+from pathlib import Path
+import sys
+
+text = Path(sys.argv[1]).read_text(encoding='utf-8')
+lines = text.splitlines()
+inside = False
+rows = []
+for line in lines:
+    if line.startswith('## 当前未修复问题速查表'):
+        inside = True
+        continue
+    if inside and line.startswith('---'):
+        break
+    if inside and line.startswith('|') and not line.startswith('| #') and not line.startswith('|---'):
+        rows.append(line)
+
+resolved = [row for row in rows if '✅ 已修复' in row or '✅ 已完成' in row or '✅ 已恢复' in row]
+print(f'BACKLOG_CURRENT_TABLE rows={len(rows)} resolved_rows={len(resolved)}')
+if resolved:
+    print('resolved rows still present in current table', file=sys.stderr)
+    raise SystemExit(1)
+PY
--- a/scripts/review/backlog_current_table_guard_test.sh
+++ b/scripts/review/backlog_current_table_guard_test.sh
@@ -0,0 +1,38 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+BAD_FILE="$TMP_DIR/bad-backlog.md"
+cat > "$BAD_FILE" <<'EOF'
+## 当前未修复问题速查表（截至 2026-05-29 12:00）
+| # | 问题 | 优先级 | 首次暴露 | 修复状态 | 影响次数 |
+|---|------|--------|----------|----------|----------|
+| 1 | old resolved item | P1 | 05-01 10:00 | ✅ 已修复 | 1 次 |
+| 2 | still open | P1 | 05-01 11:00 | ❌ 未修复 | 2 次 |
+---
+EOF
+
+set +e
+bash scripts/review/backlog_current_table_guard.sh "$BAD_FILE" >/tmp/backlog_guard_bad.out 2>&1
+BAD_RC=$?
+set -e
+[[ "$BAD_RC" -ne 0 ]]
+grep -q 'resolved rows still present in current table' /tmp/backlog_guard_bad.out
+
+GOOD_FILE="$TMP_DIR/good-backlog.md"
+cat > "$GOOD_FILE" <<'EOF'
+## 当前未修复问题速查表（截至 2026-05-29 12:00）
+| # | 问题 | 优先级 | 首次暴露 | 修复状态 | 影响次数 |
+|---|------|--------|----------|----------|----------|
+| 2 | still open | P1 | 05-01 11:00 | ❌ 未修复 | 2 次 |
+| 3 | partial item | P1 | 05-01 12:00 | ⚠️ 部分 | 3 次 |
+---
+EOF
+
+bash scripts/review/backlog_current_table_guard.sh "$GOOD_FILE" >/tmp/backlog_guard_good.out 2>&1
+grep -q 'BACKLOG_CURRENT_TABLE rows=2 resolved_rows=0' /tmp/backlog_guard_good.out
--- a/scripts/review/backlog_revocation_guard.sh
+++ b/scripts/review/backlog_revocation_guard.sh
@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+BACKLOG_PATH="${1:?backlog path required}"
+
+python3 - <<'PY' "$BACKLOG_PATH"
+from pathlib import Path
+import re
+import sys
+
+text = Path(sys.argv[1]).read_text(encoding='utf-8')
+lines = text.splitlines()
+inside = False
+current_rows = []
+for line in lines:
+    if line.startswith('## 当前未修复问题速查表'):
+        inside = True
+        continue
+    if inside and line.startswith('---'):
+        break
+    if inside and line.startswith('|') and not line.startswith('| #') and not line.startswith('|---'):
+        current_rows.append(line)
+
+resolved_ids = set(re.findall(r'#### 问题 ([0-9]+) 状态更新：已修复（从 current 表移除）', text))
+for issue_id in resolved_ids:
+    for row in current_rows:
+        if row.startswith(f'| {issue_id} |') or row.startswith(f'| {issue_id} |'.replace(' ', '')):
+            print('resolved issue still present in current table', file=sys.stderr)
+            raise SystemExit(1)
+print('BACKLOG_REVOCATION_GUARD: PASS')
+PY
--- a/scripts/review/backlog_revocation_guard_test.sh
+++ b/scripts/review/backlog_revocation_guard_test.sh
@@ -0,0 +1,40 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+BAD_BACKLOG="$TMP_DIR/backlog.md"
+cat > "$BAD_BACKLOG" <<'EOF'
+## 当前未修复问题速查表（截至 2026-05-29 15:10）
+| # | 问题 | 优先级 | 首次暴露 | 修复状态 | 影响次数 |
+|---|------|--------|----------|----------|----------|
+| 33 | 已证伪 blocker 缺少自动降级/撤销机制 | P1 | 05-18 09:30 | ❌ 未修复 | 2 次 |
+---
+#### 问题 33 状态更新：已修复（从 current 表移除）
+- 结论：问题 33 已关闭。
+EOF
+
+set +e
+bash scripts/review/backlog_revocation_guard.sh "$BAD_BACKLOG" >/tmp/backlog_revocation_bad.out 2>&1
+BAD_RC=$?
+set -e
+[[ "$BAD_RC" -ne 0 ]]
+grep -q 'resolved issue still present in current table' /tmp/backlog_revocation_bad.out
+
+GOOD_BACKLOG="$TMP_DIR/backlog-good.md"
+cat > "$GOOD_BACKLOG" <<'EOF'
+## 当前未修复问题速查表（截至 2026-05-29 15:10）
+| # | 问题 | 优先级 | 首次暴露 | 修复状态 | 影响次数 |
+|---|------|--------|----------|----------|----------|
+| 34 | 局部 smoke 已通过后缺少全局 blocker 切换提示 | P1 | 05-18 15:10 | ❌ 未修复 | 1 次 |
+---
+#### 问题 33 状态更新：已修复（从 current 表移除）
+- 结论：问题 33 已关闭。
+EOF
+
+bash scripts/review/backlog_revocation_guard.sh "$GOOD_BACKLOG" >/tmp/backlog_revocation_good.out 2>&1
+grep -q 'BACKLOG_REVOCATION_GUARD: PASS' /tmp/backlog_revocation_good.out
--- a/scripts/review/blocker_switch_guard.sh
+++ b/scripts/review/blocker_switch_guard.sh
@@ -0,0 +1,14 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+REPORT_PATH="${1:?review file required}"
+CONTENT="$(cat "$REPORT_PATH")"
+
+if [[ "$CONTENT" == *'替换'* ]]; then
+  if [[ "$CONTENT" != *'freshness_hint=same-day-blocker-switch'* ]]; then
+    echo "missing blocker switch freshness hint" >&2
+    exit 1
+  fi
+fi
+
+echo "BLOCKER_SWITCH_GUARD: PASS"
--- a/scripts/review/blocker_switch_guard_test.sh
+++ b/scripts/review/blocker_switch_guard_test.sh
@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+BAD_FILE="$TMP_DIR/review-switch.txt"
+cat > "$BAD_FILE" <<'EOF'
+### 2026-05-29 15:10（afternoon-review cron）
+- xfyun-live smoke FAIL 替换 sensenova-live
+EOF
+
+set +e
+bash scripts/review/blocker_switch_guard.sh "$BAD_FILE" >/tmp/blocker_switch_bad.out 2>&1
+BAD_RC=$?
+set -e
+[[ "$BAD_RC" -ne 0 ]]
+grep -q 'missing blocker switch freshness hint' /tmp/blocker_switch_bad.out
+
+GOOD_FILE="$TMP_DIR/review-switch-good.txt"
+cat > "$GOOD_FILE" <<'EOF'
+### 2026-05-29 15:10（afternoon-review cron）
+- xfyun-live smoke FAIL 替换 sensenova-live
+- freshness_hint=same-day-blocker-switch old=sensenova-live new=xfyun-live
+EOF
+
+bash scripts/review/blocker_switch_guard.sh "$GOOD_FILE" >/tmp/blocker_switch_good.out 2>&1
+grep -q 'BLOCKER_SWITCH_GUARD: PASS' /tmp/blocker_switch_good.out
--- a/scripts/review/current_row_revocation_guard.sh
+++ b/scripts/review/current_row_revocation_guard.sh
@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROW_FILE="${1:?row file required}"
+ROW_CONTENT="$(cat "$ROW_FILE")"
+
+if [[ "$ROW_CONTENT" == *'✅ 已修复'* || "$ROW_CONTENT" == *'✅ 已完成'* || "$ROW_CONTENT" == *'✅ 已恢复'* ]]; then
+  echo "resolved current row should be revoked from current table" >&2
+  exit 1
+fi
+
+echo "CURRENT_ROW_REVOCATION_GUARD: PASS"
--- a/scripts/review/global_blocker_switch_capture_test.sh
+++ b/scripts/review/global_blocker_switch_capture_test.sh
@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_FILE="$(mktemp)"
+trap 'rm -f "$TMP_FILE"' EXIT
+
+bash scripts/verify_phase6.sh >"$TMP_FILE" 2>&1 || true
+bash scripts/review/global_blocker_switch_guard.sh "$TMP_FILE"
+grep -q '^BLOCKER_SWITCH class=' "$TMP_FILE"
--- a/scripts/review/global_blocker_switch_guard.sh
+++ b/scripts/review/global_blocker_switch_guard.sh
@@ -0,0 +1,20 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+STATUS_FILE="${1:?phase status file required}"
+CONTENT="$(cat "$STATUS_FILE")"
+
+if [[ "$CONTENT" == *'importer_smoke_gate_result=PASS'* && "$CONTENT" == *'ROOT_CAUSE class='* && "$CONTENT" != *'ROOT_CAUSE class=none'* ]]; then
+  if [[ "$CONTENT" != *'BLOCKER_SWITCH class=global-blocker-shift'* ]]; then
+    echo "missing global blocker switch hint" >&2
+    exit 1
+  fi
+fi
+
+if [[ "$CONTENT" == *'importer_smoke_gate_result=FAIL'* && "$CONTENT" == *'live_run_result=SKIPPED'* ]]; then
+  if [[ "$CONTENT" != *'BLOCKER_SWITCH class=global-blocker-shift'* ]]; then
+    echo "missing global blocker switch hint" >&2
+    exit 1
+  fi
+fi
+echo "GLOBAL_BLOCKER_SWITCH_GUARD: PASS"
--- a/scripts/review/global_blocker_switch_guard_test.sh
+++ b/scripts/review/global_blocker_switch_guard_test.sh
@@ -0,0 +1,33 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+BAD_FILE="$TMP_DIR/phase6.txt"
+cat > "$BAD_FILE" <<'EOF'
+[PASS] importer_smoke_gate_result=PASS 新增导入器 smoke gate 通过
+[FAIL] live_run_result=FAIL 主链路真实采集失败
+ROOT_CAUSE class=primary_pipeline_failure source=live_run summary=主链路真实采集失败
+EOF
+
+set +e
+bash scripts/review/global_blocker_switch_guard.sh "$BAD_FILE" >/tmp/global_blocker_bad.out 2>&1
+BAD_RC=$?
+set -e
+[[ "$BAD_RC" -ne 0 ]]
+grep -q 'missing global blocker switch hint' /tmp/global_blocker_bad.out
+
+GOOD_FILE="$TMP_DIR/phase6-good.txt"
+cat > "$GOOD_FILE" <<'EOF'
+[PASS] importer_smoke_gate_result=PASS 新增导入器 smoke gate 通过
+[FAIL] live_run_result=FAIL 主链路真实采集失败
+ROOT_CAUSE class=primary_pipeline_failure source=live_run summary=主链路真实采集失败
+BLOCKER_SWITCH class=global-blocker-shift old=importer_smoke_gate new=live_run
+EOF
+
+bash scripts/review/global_blocker_switch_guard.sh "$GOOD_FILE" >/tmp/global_blocker_good.out 2>&1
+grep -q 'GLOBAL_BLOCKER_SWITCH_GUARD: PASS' /tmp/global_blocker_good.out
--- a/scripts/review/live_run_classification_test.sh
+++ b/scripts/review/live_run_classification_test.sh
@@ -0,0 +1,23 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+classify_live_run_failure() {
+  local live_tail="${1:-}"
+  local normalized
+  normalized="$(printf '%s' "$live_tail" | tr '[:upper:]' '[:lower:]')"
+  case "$normalized" in
+    *"api key"*|*"database_url"*|*"must provide"*|*"未设置"*|*"permission denied"*|*"role does not exist"*|*"relation does not exist"*)
+      printf '%s\n' 'precondition_missing'
+      ;;
+    *"signature_guard"*|*"unexpected status 403"*|*"unexpected status 502"*|*"unexpected status 503"*|*"unexpected status 504"*|*"no pricing cards found"*|*"no model rows parsed"*|*"no model overview cards parsed"*|*"context deadline exceeded"*|*"client.timeout"*|*"i/o timeout"*|*"tls handshake timeout"*|*"transport closed"*|*"connection reset"*|*"connection refused"*|*"no such host"*)
+      printf '%s\n' 'external_provider_failure'
+      ;;
+    *)
+      printf '%s\n' 'primary_pipeline_failure'
+      ;;
+  esac
+}
+
+[[ "$(classify_live_run_failure 'OPENROUTER_API_KEY 未设置，无法执行真实采集')" == 'precondition_missing' ]]
+[[ "$(classify_live_run_failure 'import_vertex_pricing: read https://cloud.google.com/...: context deadline exceeded (Client.Timeout or context cancellation while reading body)')" == 'external_provider_failure' ]]
+[[ "$(classify_live_run_failure 'insert report_runs failed: duplicate key value violates unique constraint')" == 'primary_pipeline_failure' ]]
--- a/scripts/review/partial_output_guard.sh
+++ b/scripts/review/partial_output_guard.sh
@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+STATUS_FILE="${1:?status file required}"
+CONTENT="$(cat "$STATUS_FILE")"
+
+if [[ "$CONTENT" != *'PARTIAL_OUTPUT_STATUS complete=false source=artifact-required conclusion=conservative'* ]]; then
+  echo "missing conservative partial-output template" >&2
+  exit 1
+fi
+
+echo "PARTIAL_OUTPUT_GUARD: PASS"
--- a/scripts/review/partial_output_guard_test.sh
+++ b/scripts/review/partial_output_guard_test.sh
@@ -0,0 +1,30 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+BAD_FILE="$TMP_DIR/bad-output.txt"
+cat > "$BAD_FILE" <<'EOF'
+long command output line 1
+long command output line 2
+EOF
+
+set +e
+bash scripts/review/partial_output_guard.sh "$BAD_FILE" >/tmp/partial_bad.out 2>&1
+BAD_RC=$?
+set -e
+[[ "$BAD_RC" -ne 0 ]]
+grep -q 'missing conservative partial-output template' /tmp/partial_bad.out
+
+GOOD_FILE="$TMP_DIR/good-output.txt"
+cat > "$GOOD_FILE" <<'EOF'
+PARTIAL_OUTPUT_STATUS complete=false source=artifact-required conclusion=conservative
+artifact_hint=artifact://123
+EOF
+
+bash scripts/review/partial_output_guard.sh "$GOOD_FILE" >/tmp/partial_good.out 2>&1
+grep -q 'PARTIAL_OUTPUT_GUARD: PASS' /tmp/partial_good.out
--- a/scripts/review/provider_root_cause_test.sh
+++ b/scripts/review/provider_root_cause_test.sh
@@ -0,0 +1,20 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+classify_live_run_provider() {
+  local live_tail="${1:-}"
+  local normalized
+  normalized="$(printf '%s' "$live_tail" | tr '[:upper:]' '[:lower:]')"
+  case "$normalized" in
+    *"import_vertex_pricing"*) printf '%s\n' 'vertex_pricing' ;;
+    *"import_cloudflare_pricing"*|*"cloudflare_pricing"*) printf '%s\n' 'cloudflare_pricing' ;;
+    *"import_perplexity_pricing"*|*"perplexity_pricing"*) printf '%s\n' 'perplexity_pricing' ;;
+    *"import_xfyun_pricing"*|*"xfyun_pricing"*) printf '%s\n' 'xfyun_pricing' ;;
+    *) printf '%s\n' 'unknown_external_provider' ;;
+  esac
+}
+
+[[ "$(classify_live_run_provider 'import_vertex_pricing: read https://cloud.google.com/...: context deadline exceeded')" == 'vertex_pricing' ]]
+[[ "$(classify_live_run_provider 'import_cloudflare_pricing: fetch https://developers.cloudflare.com/...: unexpected status 403')" == 'cloudflare_pricing' ]]
+[[ "$(classify_live_run_provider 'import_perplexity_pricing: no model rows parsed')" == 'perplexity_pricing' ]]
+[[ "$(classify_live_run_provider 'import_xfyun_pricing: no pricing cards found')" == 'xfyun_pricing' ]]
--- a/scripts/review/release_semantics_capture_test.sh
+++ b/scripts/review/release_semantics_capture_test.sh
@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_FILE="$(mktemp)"
+trap 'rm -f "$TMP_FILE"' EXIT
+
+bash scripts/verify_phase6.sh >"$TMP_FILE" 2>&1 || true
+bash scripts/review/release_semantics_guard.sh "$TMP_FILE"
+grep -q '^RELEASE_SEMANTICS class=' "$TMP_FILE"
--- a/scripts/review/release_semantics_guard.sh
+++ b/scripts/review/release_semantics_guard.sh
@@ -0,0 +1,11 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+STATUS_FILE="${1:?phase status file required}"
+
+if ! grep -q '^RELEASE_SEMANTICS class=' "$STATUS_FILE"; then
+  echo "missing release semantics summary" >&2
+  exit 1
+fi
+
+echo "RELEASE_SEMANTICS_GUARD: PASS"
--- a/scripts/review/release_semantics_guard_test.sh
+++ b/scripts/review/release_semantics_guard_test.sh
@@ -0,0 +1,29 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+BAD_FILE="$TMP_DIR/bad-phase6.txt"
+cat > "$BAD_FILE" <<'EOF'
+[WARN] window_gate_result=WARN 最近 7 次采集成功率未达 95%（仅外部文档站失败：external_provider_failure_only；stability_label=recovered-external-incident）
+EOF
+
+set +e
+bash scripts/review/release_semantics_guard.sh "$BAD_FILE" >/tmp/release_semantics_bad.out 2>&1
+BAD_RC=$?
+set -e
+[[ "$BAD_RC" -ne 0 ]]
+grep -q 'missing release semantics summary' /tmp/release_semantics_bad.out
+
+GOOD_FILE="$TMP_DIR/good-phase6.txt"
+cat > "$GOOD_FILE" <<'EOF'
+[WARN] window_gate_result=WARN 最近 7 次采集成功率未达 95%（仅外部文档站失败：external_provider_failure_only；stability_label=recovered-external-incident）
+RELEASE_SEMANTICS class=degraded-external-provider gate=window_gate policy=release-allowed-with-warning
+EOF
+
+bash scripts/review/release_semantics_guard.sh "$GOOD_FILE" >/tmp/release_semantics_good.out 2>&1
+grep -q 'RELEASE_SEMANTICS_GUARD: PASS' /tmp/release_semantics_good.out
--- a/scripts/review/review_action_guard.sh
+++ b/scripts/review/review_action_guard.sh
@@ -0,0 +1,35 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+REPORT_PATH="${1:?review report path required}"
+
+NEXT_BLOCK="$(python3 - <<'PY' "$REPORT_PATH"
+from pathlib import Path
+import sys
+text = Path(sys.argv[1]).read_text(encoding='utf-8')
+lines = text.splitlines()
+inside = False
+buf = []
+for line in lines:
+    if line.startswith('## '):
+        if inside:
+            break
+        inside = line.strip() == '## Next'
+        continue
+    if inside:
+        buf.append(line)
+print('\n'.join(buf))
+PY
+)"
+
+if [[ -z "${NEXT_BLOCK//[[:space:]]/}" ]]; then
+  echo "missing actionable next step: ## Next block is empty" >&2
+  exit 1
+fi
+
+if ! printf '%s\n' "$NEXT_BLOCK" | grep -Eq '(^- .*处理|^- .*修复|^- .*新增|^- .*更新|^- .*验证|^- .*运行|^- .*同步|^- .*接入|^- .*排查|^[0-9]+\..*(处理|修复|新增|更新|验证|运行|同步|接入|排查)|处理问题|修复问题|新增.*guard|接入.*链)'; then
+  echo "missing actionable next step: ## Next does not contain executable action" >&2
+  exit 1
+fi
+
+echo "REVIEW_ACTION_GUARD: PASS"
--- a/scripts/review/review_action_guard_test.sh
+++ b/scripts/review/review_action_guard_test.sh
@@ -0,0 +1,47 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+BAD_REPORT="$TMP_DIR/bad-review.md"
+cat > "$BAD_REPORT" <<'EOF'
+## Context
+- no delta
+
+## Evidence
+- one finding
+
+## Outcome
+- still blocked
+
+## Next
+- keep watching
+EOF
+
+set +e
+bash scripts/review/review_action_guard.sh "$BAD_REPORT" >/tmp/review_action_bad.out 2>&1
+BAD_RC=$?
+set -e
+[[ "$BAD_RC" -ne 0 ]]
+grep -q 'missing actionable next step' /tmp/review_action_bad.out
+
+GOOD_REPORT="$TMP_DIR/good-review.md"
+cat > "$GOOD_REPORT" <<'EOF'
+## Context
+- no delta
+
+## Evidence
+- one finding
+
+## Outcome
+- still blocked
+
+## Next
+- 处理问题 11，新增 review action guard 并接入 review 生成链
+EOF
+
+bash scripts/review/review_action_guard.sh "$GOOD_REPORT" >/tmp/review_action_good.out 2>&1
--- a/scripts/review/review_aging_priority_test.sh
+++ b/scripts/review/review_aging_priority_test.sh
@@ -0,0 +1,20 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+cat > "$TMP_DIR/git_status.txt" <<'EOF'
+EOF
+
+cat > "$TMP_DIR/backlog_rows.txt" <<'EOF'
+| 18 | no delta aging | P2 | 05-12 22:46 | ❌ 未修复 | 7 次 |
+| 14 | phase6plus undefined | P1 | 05-10 21:30 | ❌ 未修复 | 6 次 |
+| 15 | review false positive | P1 | 05-11 14:30 | ❌ 未修复 | 10 次 |
+EOF
+
+OUTPUT="$(bash scripts/review/review_status_summary.sh "$TMP_DIR/git_status.txt" "$TMP_DIR/backlog_rows.txt")"
+printf '%s' "$OUTPUT" | grep -q 'aging_focus=14:P1:6,15:P1:10,18:P2:7'
--- a/scripts/review/review_same_day_no_decision_test.sh
+++ b/scripts/review/review_same_day_no_decision_test.sh
@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+cat > "$TMP_DIR/git_status.txt" <<'EOF'
+EOF
+
+cat > "$TMP_DIR/backlog_rows.txt" <<'EOF'
+| 31 | no decision delta | P2 | 05-17 15:10 | ❌ 未修复 | 3 次 |
+| 30 | aged precondition | P1 | 05-17 09:31 | ❌ 未修复 | 6 次 |
+EOF
+
+set +e
+OUTPUT="$(bash scripts/review/review_status_summary.sh "$TMP_DIR/git_status.txt" "$TMP_DIR/backlog_rows.txt")"
+set -e
+printf '%s' "$OUTPUT" | grep -q 'same_day_no_decision_focus='
+printf '%s' "$OUTPUT" | grep -q 'same_day_no_decision_focus=30:P1:6,31:P2:3'
--- a/scripts/review/review_status_summary.sh
+++ b/scripts/review/review_status_summary.sh
@@ -0,0 +1,69 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+GIT_STATUS_PATH="${1:?git status file required}"
+BACKLOG_ROWS_PATH="${2:?backlog rows file required}"
+
+DIRTY_WORKTREE=0
+if [[ -s "$GIT_STATUS_PATH" ]]; then
+  DIRTY_WORKTREE=1
+fi
+
+OPEN_ISSUES=$(grep -c '^|' "$BACKLOG_ROWS_PATH" || true)
+NO_DELTA=true
+FOCUS="risk_aging,unverified,backlog"
+AGING_FOCUS="$(python3 - <<'PY' "$BACKLOG_ROWS_PATH"
+from pathlib import Path
+import sys
+rows=[]
+for line in Path(sys.argv[1]).read_text(encoding='utf-8').splitlines():
+    if not line.startswith('|'):
+        continue
+    parts=[p.strip() for p in line.strip('|').split('|')]
+    if len(parts) < 6:
+        continue
+    issue_id=parts[0]
+    priority=parts[2]
+    first_exposed=parts[3]
+    impact=parts[5].replace('次','').strip()
+    try:
+        impact_num=int(''.join(ch for ch in impact if ch.isdigit()) or '0')
+    except ValueError:
+        impact_num=0
+    rows.append((priority, first_exposed, -impact_num, issue_id, impact_num))
+priority_rank={'P0':0,'P1':1,'P2':2}
+rows.sort(key=lambda x:(priority_rank.get(x[0],9), x[1], x[2], x[3]))
+focus=[]
+for priority, first_exposed, neg_impact, issue_id, impact_num in rows[:3]:
+    focus.append(f"{issue_id}:{priority}:{impact_num}")
+print(','.join(focus))
+PY
+)"
+SAME_DAY_NO_DECISION_FOCUS="$(python3 - <<'PY' "$BACKLOG_ROWS_PATH"
+from pathlib import Path
+import sys
+rows=[]
+for line in Path(sys.argv[1]).read_text(encoding='utf-8').splitlines():
+    if not line.startswith('|'):
+        continue
+    parts=[p.strip() for p in line.strip('|').split('|')]
+    if len(parts) < 6:
+        continue
+    issue_id=parts[0]
+    priority=parts[2]
+    impact=parts[5].replace('次','').strip()
+    try:
+        impact_num=int(''.join(ch for ch in impact if ch.isdigit()) or '0')
+    except ValueError:
+        impact_num=0
+    rows.append((priority, -impact_num, issue_id, impact_num))
+priority_rank={'P0':0,'P1':1,'P2':2}
+rows.sort(key=lambda x:(priority_rank.get(x[0],9), x[1], x[2]))
+focus=[]
+for priority, neg_impact, issue_id, impact_num in rows[:2]:
+    focus.append(f"{issue_id}:{priority}:{impact_num}")
+print(','.join(focus))
+PY
+)"
+
+printf 'REVIEW_STATUS no_delta=%s dirty_worktree=%d open_issues=%d focus=%s aging_focus=%s same_day_no_decision_focus=%s\n' "$NO_DELTA" "$DIRTY_WORKTREE" "$OPEN_ISSUES" "$FOCUS" "$AGING_FOCUS" "$SAME_DAY_NO_DECISION_FOCUS"
--- a/scripts/review/review_status_summary_test.sh
+++ b/scripts/review/review_status_summary_test.sh
@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+cat > "$TMP_DIR/git_status.txt" <<'EOF'
+EOF
+
+cat > "$TMP_DIR/backlog_rows.txt" <<'EOF'
+| 8 | cron review 无 delta 时空转 | P1 | 05-08 09:12 | ❌ 未修复 | 13 次 |
+| 9 | 验证模式伪进展（artifact_present 局限） | P1 | 05-08 14:30 | ❌ 未修复 | 10 次 |
+EOF
+
+OUTPUT="$(bash scripts/review/review_status_summary.sh "$TMP_DIR/git_status.txt" "$TMP_DIR/backlog_rows.txt")"
+printf '%s' "$OUTPUT" | grep -q 'REVIEW_STATUS no_delta=true'
+printf '%s' "$OUTPUT" | grep -q 'dirty_worktree=0'
+printf '%s' "$OUTPUT" | grep -q 'open_issues=2'
+printf '%s' "$OUTPUT" | grep -q 'focus=risk_aging,unverified,backlog'
--- a/scripts/review/review_truth_guard.sh
+++ b/scripts/review/review_truth_guard.sh
@@ -0,0 +1,24 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+REPORT_PATH="${1:?review report path required}"
+CONTENT="$(cat "$REPORT_PATH")"
+
+if [[ "$CONTENT" != *'runtime-verified'* ]]; then
+  echo "missing truth label: runtime-verified" >&2
+  exit 1
+fi
+
+if [[ "$CONTENT" != *'current status'* && "$CONTENT" != *'当前状态'* ]]; then
+  echo "missing truth label: current status" >&2
+  exit 1
+fi
+
+if [[ "$CONTENT" == *'历史状态'* || "$CONTENT" == *'historical status'* || "$CONTENT" == *'历史 review'* || "$CONTENT" == *'上一轮 review'* ]]; then
+  if [[ "$CONTENT" != *'历史状态'* && "$CONTENT" != *'historical status'* ]]; then
+    echo "missing truth label: historical status" >&2
+    exit 1
+  fi
+fi
+
+echo "REVIEW_TRUTH_GUARD: PASS"
--- a/scripts/review/review_truth_guard_test.sh
+++ b/scripts/review/review_truth_guard_test.sh
@@ -0,0 +1,62 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "$ROOT_DIR"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+BAD_REPORT="$TMP_DIR/bad-review.md"
+cat > "$BAD_REPORT" <<'EOF'
+## Evidence
+### Incomplete
+- 未完成项：外部文档源仍不稳定
+  - 当前状态：已经恢复绿色
+EOF
+
+set +e
+bash scripts/review/review_truth_guard.sh "$BAD_REPORT" >/tmp/review_truth_bad.out 2>&1
+BAD_RC=$?
+set -e
+[[ "$BAD_RC" -ne 0 ]]
+grep -q 'missing truth label' /tmp/review_truth_bad.out
+
+GOOD_REPORT="$TMP_DIR/good-review.md"
+cat > "$GOOD_REPORT" <<'EOF'
+## Evidence
+### Incomplete
+- 未完成项：外部文档源仍不稳定
+  - 当前状态：`runtime-verified`；最近一轮 `verify_phase6.sh` 已通过，但此条仅代表当前快照，不代表历史报告中的旧 FAIL 仍然成立。
+EOF
+
+bash scripts/review/review_truth_guard.sh "$GOOD_REPORT" >/tmp/review_truth_good.out 2>&1
+
+DRIFT_REPORT="$TMP_DIR/drift-review.md"
+cat > "$DRIFT_REPORT" <<'EOF'
+## Evidence
+### Inconsistencies
+- 伪进展或文档/实现不一致项：历史 review 仍保留旧 FAIL
+  - 证据：上一轮 review 写了 FAIL
+  - 当前状态：`runtime-verified`；当前 verify_phase6 已通过
+EOF
+
+set +e
+bash scripts/review/review_truth_guard.sh "$DRIFT_REPORT" >/tmp/review_truth_drift.out 2>&1
+DRIFT_RC=$?
+set -e
+[[ "$DRIFT_RC" -ne 0 ]]
+grep -q 'missing truth label: historical status' /tmp/review_truth_drift.out
+
+GOOD_DRIFT_REPORT="$TMP_DIR/good-drift-review.md"
+cat > "$GOOD_DRIFT_REPORT" <<'EOF'
+## Evidence
+### Inconsistencies
+- 伪进展或文档/实现不一致项：历史 review 仍保留旧 FAIL
+  - 历史状态：上一轮 review 记录为 FAIL
+  - 当前状态：`runtime-verified`；当前 verify_phase6 已通过
+EOF
+
+bash scripts/review/review_truth_guard.sh "$GOOD_DRIFT_REPORT" >/tmp/review_truth_good_drift.out 2>&1
+grep -q 'REVIEW_TRUTH_GUARD: PASS' /tmp/review_truth_good_drift.out
+grep -q 'REVIEW_TRUTH_GUARD: PASS' /tmp/review_truth_good.out
--- a/Show More
+++ b/Show More