coding-agent-harness 1.0.1 → 1.0.2

package/references/legacy-12-phase-bootstrap.md

@@ -0,0 +1,41 @@
+# Historical 12-Phase Bootstrap
+
+This document preserves the legacy bootstrap sequence used before the v1.0 capability-based installer. It is reference material for migration only. Agents should not use this as the default install protocol for new projects.
+
+## When To Read
+
+- You are migrating a project that was created with the older 12-phase harness flow.
+- You need to map older documents into v1.0 capabilities and task contracts.
+- You need to understand why a legacy project has broad reference files, embedded roadmap sections, or single-line phase planning.
+
+## Legacy Sequence
+
+1. Project diagnosis: read `references/project-onboarding-audit.md` and produce a project diagnosis.
+2. Solution confirmation: decide harness scale from the diagnosis.
+3. Delivery operating model: classify solo, team, split-repo, program, stage-gate, or kanban delivery.
+4. Optional module registration: identify independent modules, write scopes, module registry rows, and module plans.
+5. Directory structure: create or adapt `docs/` according to the docs directory standard.
+6. AGENTS.md and CLAUDE.md: generate the root agent entrypoint and Claude Code shim.
+7. Reference standards: generate project-specific reference files under `docs/11-REFERENCE/`.
+8. Repository governance and CI/CD: define PR policy, required checks, branch protection, CI, and worktree concurrency.
+9. Planning loop: create task templates and task directories.
+10. Long-running task protocol: add long-running task standard and contract template.
+11. SSoT, lessons, harness ledger, regression, cadence, walkthrough, and closeout files.
+12. Bootstrap summary: report created files, purpose, first tasks, next actions, and checker status.
+
+## v1.0 Mapping
+
+| Legacy Area | v1.0 Destination |
+| --- | --- |
+| AGENTS.md / CLAUDE.md | Preserve existing files; merge only missing routing and residuals. |
+| Task directory with only `task_plan.md` | Add `brief.md`, `execution_strategy.md`, `visual_map.md`, `progress.md`, `findings.md`, and `review.md` only for active tasks. |
+| Embedded roadmap in `task_plan.md` | Move active roadmap rows into standalone `visual_map.md`; leave historical tasks untouched unless they are reopened. |
+| Single-line progress status | Normalize active tasks through `harness task-start`, `task-block`, `task-log`, and `task-complete`. |
+| Broad reference bundle | Declare only capabilities that are actually adopted in `.harness-capabilities.json`. |
+| Long-running task artifacts | Add or declare `long-running-task` only when active work needs continuous autonomous execution. |
+| Informal module lists | Adopt `module-parallel` only after modules have owners, write scopes, dependency rules, and registry maintenance. |
+| Historical review notes | Do not rewrite all old review files. Upgrade active release-blocking reviews to the v1 review schema first. |
+
+## Migration Rule
+
+Do not mechanically rewrite the whole project. Use `harness add-capability safe-adoption` to add the v1.0 compatibility layer, then use `harness migrate-plan --json` to produce a staged action list. Migrate active or reopened work first; leave closed historical tasks as legacy evidence unless strict gates require them.

package/references/lessons-governance.md

@@ -43,28 +43,45 @@ docs/01-GOVERNANCE/
 3. 有没有踩坑经验值得记录，避免下次重复？
 4. 有没有架构层面的洞察，值得更新架构文档？
 
-如果任何一条答案是"有"，必须执行“双写”：
+如果任何一条答案是"有"，先写入任务目录的 `lesson_candidates.md`，由人工 review 决定是否进入治理 promotion。
+人工决定后只允许以下任务级状态：
+
+- `no-candidate-accepted`：接受本轮没有可复用 lesson。
+- `needs-promotion`：至少一个候选已排队进入治理沉淀。
+- `promoted`：维护命令已经写入 Lessons SSoT 和详情文档。
+- `rejected`：候选已带理由拒绝。
+
+`needs-promotion` 不阻塞任务 closeout，但必须在 Closeout SSoT / Harness Ledger 中记录
+`queued-promotion: LC-YYYYMMDD-NNN`，并由后续维护命令处理。`promoted` 或人工直接创建时记录
+`checked-created: L-YYYY-MM-DD-NNN`。如果没有候选，记录 `checked-candidate: LC-...` 或
+`checked-none: <reason>`，其中 `checked-none` 只用于旧任务兼容或没有 candidate 文件的历史收口。
+
+promotion 执行“双写”：
 
 1. 选择 `templates/lessons/` 下的对应模板，先创建详情文档。
 2. 详情文档必须写清背景、现状问题、建议改动、影响范围和冲突声明。
 3. 再在 Lessons SSoT 追加表行，`Detail Doc` 指向这篇详情文档。
-4. 在 Closeout SSoT / Harness Ledger 中记录 `checked-created`，并引用 lesson ID。
+4. 在 Closeout SSoT / Harness Ledger 中记录 `checked-created` 或 `queued-promotion`，并引用 lesson/candidate ID。
 
-如果四个问题的答案全是"没有"，也不能静默跳过。必须在 Closeout SSoT 和
-Harness Ledger 中记录 `checked-none: <一句话原因>`。
+如果四个问题的答案全是"没有"，也不能静默跳过。新任务必须在 `lesson_candidates.md` 中使用
+`no-candidate-accepted` 并填写 No-Candidate Reason；旧任务可在 Closeout SSoT 和 Harness Ledger 中记录
+`checked-none: <一句话原因>`。
 
 ## Closeout 判定
 
-收口时只允许以下两种合格状态：
+收口时只允许以下合格状态：
 
 - `checked-created: L-YYYY-MM-DD-NNN`：发现可沉淀经验，已创建详情文档并更新 Lessons SSoT。
-- `checked-none: <reason>`：已完整检查，确认本轮没有可复用的规范/流程/架构改进。
+- `queued-promotion: LC-YYYYMMDD-NNN`：人工确认候选值得沉淀，但交给维护命令后续提升。
+- `checked-candidate: LC-YYYYMMDD-NNN`：人工已审查 candidate 文件，结论为无候选或全部拒绝。
+- `checked-none: <reason>`：旧任务兼容状态，已完整检查且没有 candidate 文件。
 
 以下状态不合格：
 
 - 只写 Lessons SSoT 表行，没有详情文档。
 - 只写详情文档，没有 Lessons SSoT 表行。
 - 在 walkthrough 或 progress 中说“无 lessons”，但 Closeout SSoT / Harness Ledger 没有记录。
+- 新任务跳过 `lesson_candidates.md`，只用 `checked-none` 代替 candidate 判定。
 - 用 `n/a` 代替检查结果，除非任务是纯只读分析且没有 closed ledger row。
 
 ## 沉淀类型

package/references/planning-loop.md

@@ -6,11 +6,21 @@
 
 ## 任务目录结构
 
+任务信息架构预算决定默认脚手架：
+
+| Budget | 适用场景 | 默认文件 |
+| --- | --- | --- |
+| `simple` | 单 owner、无 subagent、L0/L1 证据、可跳过正式 review gate | `brief.md`, `task_plan.md`, `visual_map.md`, `progress.md` |
+| `standard` | 常规功能、修复、文档任务，需要完整可追溯记录 | 完整任务文件 |
+| `complex` | L2/L3 证据、subagent/reviewer、外部参考、生成产物或 optional indexes | 完整任务文件 + 按需 optional structure |
+
+`trivial` 不进入 CLI：小到不值得建立任务目录的修改，可以直接执行并在 commit 或交付说明中写清楚原因。
+
 ```
 docs/09-PLANNING/TASKS/<YYYY-MM-DD-任务名>/
 ├── task_plan.md    ← 计划：目标、范围、步骤、验收标准
 ├── execution_strategy.md ← 执行策略：模式、subagent、冲突控制、验证深度、handoff
-├── visual_roadmap.md ← 可视化路线：Mermaid、phase table、完成度、证据状态
+├── visual_map.md ← 图表集合：阶段图、架构图、时序图、数据流、状态机、完成度、证据状态
 ├── findings.md     ← 发现：执行过程中的研究发现和技术决策
 ├── progress.md     ← 进度：每个阶段的状态更新和验证结果
 ├── review.md       ← 对抗性审查报告（需要 reviewer / subagent / release review 时必填）
@@ -43,8 +53,9 @@ slices/<slice-id>/review.md
 6. **长程任务必须补合同** — 如果任务需要连续执行、多轮审查或子代理 review，先补 `long-running-task-contract.md`
 7. **任务收口必须回写 Harness Ledger** — 只在任务完成或上下文回写状态变化时记录，不记录每次 `progress.md` 更新
 8. **复杂任务必须记录 `execution_strategy.md`** — 是否使用 subagent、reviewer、worktree、handoff 都写入独立文件。
-9. **非平凡任务必须记录 `visual_roadmap.md`** — HTML dashboard 从独立文件的 Mermaid 和 phase table 计算完成度、阻塞和证据状态。
+9. **非平凡任务必须记录 `visual_map.md`** — 这是任务图表集合，不只是 roadmap；HTML dashboard 从独立文件的 phase table 计算完成度、阻塞和证据状态。
 10. **路径必须带来源前缀** — 使用 `PUBLIC:`, `PRIVATE:`, `TARGET:`, `EXTERNAL:`, `URL:`，避免脆弱相对路径。
+11. **已验证切片必须主动提交** — 每个有意义的工作切片通过对应检查后，agent 默认主动 commit；只有用户明确要求暂不提交、检查失败、或 dirty diff 归属未清时才延期，并把原因写入 progress / handoff。
 
 ## task_plan.md 模板
 
@@ -68,7 +79,7 @@ slices/<slice-id>/review.md
 | Contract File | Required | Purpose |
 | --- | --- | --- |
 | `execution_strategy.md` | yes | Execution mode, subagent use, conflict control, evidence depth, handoff rules |
-| `visual_roadmap.md` | yes | Mermaid route, phase table, completion, evidence status, blocking risk |
+| `visual_map.md` | yes | Mermaid maps, phase table, architecture/sequence/data-flow/state diagrams when useful, completion, evidence status, blocking risk |
 
 ## 步骤
 1. [步骤1]
@@ -181,12 +192,39 @@ Evidence values use `type:path:summary`.
 ## 状态流转
 
 ```
+simple:
 未开始 → 进行中 → 已完成
               ↓
           已阻塞 → 进行中
+
+standard / complex:
+未开始 → 进行中 → 审查中 → 已完成
+              ↓
+          已阻塞 → 进行中
 ```
 
+`task-review` 是 standard / complex 任务进入执行审查的唯一 CLI 路径。`task-complete`
+对 standard / complex 是硬门禁：当前状态不是 `review` 时必须拒绝。`simple`
+可以直接从 `in_progress` 完成。
+
 每次状态变更时，必须同时更新 progress.md 和 Feature SSoT。
 
 任务完成时，必须在 `docs/Harness-Ledger.md` 中记录本轮 task plan、SSoT、
 walkthrough、Lessons 检查等上下文回写结果。
+
+## Commit Convention
+
+非平凡任务不是等用户提醒才提交。每个已验证、有意义、范围清晰的切片都应形成 commit。
+提交前只 stage 本任务范围内文件；无关 dirty 文件、私有文件和生成产物必须保留原样或按项目规则处理。
+
+任务相关 commit 应在 message footer 中引用任务 ID：
+
+```text
+feat: implement task review gate
+
+Harness: TASKS/2026-05-21-task-review-gate
+```
+
+格式：`Harness: <task-id>`，其中 task-id 是 `task-list --json` 输出的 `id` 字段。
+不建目录的小修改可以使用 `Harness: trivial` 或省略 footer。1.0 只定义约定，不强制
+扫描 git 历史；后续可通过 reconcile 命令补工具化检查。

package/references/project-onboarding-audit.md

@@ -25,6 +25,8 @@
 - 有没有 docs/ 目录？结构如何？
 - 有没有 README 以外的开发文档？
 - 有没有架构设计文档？
+- 如果项目属于微服务、多仓或前后端分仓，用户是否有外部服务文档、接口文档、架构图、会议纪要、链接或导出包？
+- 外部资料能否复制进本仓？如果不能，是否需要只保存本地路径、URL、owner 和摘要？
 
 ### 4. 现有测试
 - 有没有测试框架？（jest / vitest / pytest / go test 等）
@@ -98,6 +100,7 @@
 | Regression 体系 | [有/无/需改造] | 🟢/🟡/🔴 |
 | CI/CD | [有/无/需改造] | 🟢/🟡/🔴 |
 | Repo Governance | [有/无/需改造] | 🟢/🟡/🔴 |
+| External Source Intake | [无外部资料 / 少量 evidence 链接 / 需要 source pack / 需询问用户] | 🟢/🟡/🔴 |
 | Branch Protection | [designed/implemented/verified/blocked] | 🟢/🟡/🔴 |
 | Required Checks | [有/无/需改造] | 🟢/🟡/🔴 |
 | Harness Ledger | [有/无/需改造] | 🟢/🟡/🔴 |
@@ -109,6 +112,13 @@
 2. [Surface 2]：[描述]
 ...
 
+## 外部资料判断
+
+- 是否属于微服务 / 多仓 / 前后端分仓：[是/否/不确定]
+- 是否已询问用户外部资料：[是/否/不适用]
+- 外部资料规模：[无 / 1-4 份 / 5+ 份 / 未知]
+- 推荐处理：[仅作为 Source Evidence / 创建 external-source-packs/<source-key> / 暂停等待用户提供资料 / 不适用]
+
 ## 推荐 Harness 规模
 [Lite / Standard / Full]（见下方项目类型分支）
 

package/references/repo-governance-standard.md

@@ -91,6 +91,7 @@ Required fields:
 - Branch pattern
 - Ownership rule
 - Subagent worker rule: each code-changing worker uses its own worktree / branch and hands off a commit SHA
+- Checkpoint commit rule: verified, meaningful slices are committed proactively; deferred commits require an explicit reason
 - Merge ordering rule
 - Cleanup rule
 
@@ -129,3 +130,4 @@ Bootstrap is not complete unless repository governance is at least:
 - Branch protection: `designed` with residual, or `verified`
 - Worktree concurrency: `implemented`
 - Harness checker: passing or blocked-with-owner with explicit residual
+- Checkpoint commit rule: `implemented`

package/references/testing-standard.md

@@ -0,0 +1,50 @@
+# Testing Standard
+
+## Purpose
+
+Define the minimum verification required before an agent claims that a change is complete.
+
+## Minimum Evidence by Change Type
+
+| Change Type | Minimum Evidence |
+| --- | --- |
+| Bug fix | Focused regression test or a verification command that would have exposed the bug |
+| Parser, lifecycle, scanner, or protocol behavior | Unit test coverage is required; prefer TDD |
+| CLI behavior | CLI assertion for success and failure paths |
+| Dashboard or UI behavior | Dashboard smoke plus browser/workbench interaction evidence when behavior changes |
+| Documentation, configuration, or research | Verification command or explicit waiver with residual risk |
+| High-risk shared code | TDD is expected; regression assertion is required |
+
+## TDD Policy
+
+Use TDD for behavior changes, bug fixes, lifecycle transitions, parsers, scanners, and shared protocol logic:
+
+1. Add or update a failing test that describes the desired behavior.
+2. Run the test and confirm it fails for the expected reason.
+3. Implement the smallest change that makes the test pass.
+4. Run the full relevant verification set before closeout.
+
+If TDD is skipped, record why the change is not testable or why a smoke/manual check is the better evidence.
+
+## Rules
+
+1. Tests must match the change risk. Use unit tests for isolated logic, integration tests for contracts, end-to-end or smoke tests for user workflows, and manual verification for surfaces automation cannot cover.
+2. Do not claim success from tests that did not exercise the changed behavior.
+3. A skipped, flaky, or unavailable test is residual risk and must be recorded with a reason.
+4. When fixing a bug, add or update coverage that would have caught the failure unless the cost is clearly unjustified and documented.
+5. For UI changes, verify realistic viewport and interaction states when layout, navigation, or user workflow changed.
+6. For data, migration, security, or deployment changes, include evidence from the real contract or environment where feasible.
+7. Final summaries must distinguish checks run, checks not run, and evidence inspected.
+
+## Required Checklist
+
+- Changed behavior is mapped to a verification method.
+- Required checks were run from the correct branch or worktree.
+- Test output, logs, screenshots, or command results are captured or summarized.
+- Manual checks include exact path and expected result.
+- Known gaps and residual risk are recorded.
+- Regression SSoT is updated when the change adds or changes a protected surface.
+
+## Closeout Expectations
+
+Testing closeout must list commands or manual checks, summarize results, identify untested risk, and explain why the evidence is sufficient for the requested acceptance criteria.

package/references/walkthrough-closeout.md

@@ -35,8 +35,8 @@
 - 本轮有没有发现 reference / workflow / checker 不够用或有误：[有/无，写一句理由]
 - 有没有反复出现、跨页面/跨模块/跨阶段的共性问题：[有/无，写一句理由]
 - 有没有下次 agent 也可能重复踩的坑：[有/无，写一句理由]
-- Lessons 结果：[checked-created: L-YYYY-MM-DD-NNN / checked-none: 一句话原因]
-- Lessons Detail Doc：[如 checked-created，填 `docs/01-GOVERNANCE/lessons/...md`；否则写"无"]
+- Lessons 结果：[checked-created: L-YYYY-MM-DD-NNN / queued-promotion: LC-YYYYMMDD-NNN / checked-candidate: LC-YYYYMMDD-NNN / checked-none: 一句话原因]
+- Lessons Detail Doc：[如 checked-created，填 `docs/01-GOVERNANCE/lessons/...md`；如 queued/checked-candidate，填 `lesson_candidates.md`；否则写"无"]
 
 ## 相关文件
 - Task Plan: [路径]
@@ -108,11 +108,12 @@ docs/10-WALKTHROUGH/Closeout-SSoT.md
 
 1. 完整读一遍 `docs/01-GOVERNANCE/Lessons-SSoT.md`
 2. 按 `references/lessons-governance.md` 中的规则处理冲突
-3. 在 `docs/01-GOVERNANCE/lessons/` 下写入详细建议（使用 `templates/lessons/` 下的对应模板）
-4. 更新 Lessons SSoT 表，`Detail Doc` 必须指向刚写的详情文档
-5. 在 Closeout SSoT 和 Harness Ledger 中记录 `checked-created: L-YYYY-MM-DD-NNN`
+3. 先在任务目录 `lesson_candidates.md` 中登记候选并等待人工 review
+4. 人工确认后，如需沉淀，使用 maintenance CLI 写入 `docs/01-GOVERNANCE/lessons/` 和 Lessons SSoT
+5. 在 Closeout SSoT 和 Harness Ledger 中记录 `queued-promotion: LC-...` 或 `checked-created: L-YYYY-MM-DD-NNN`
 
 如果所有答案都是“没有”，不能静默跳过；在 Closeout SSoT 和 Harness Ledger 中记录
+新任务在 `lesson_candidates.md` 中写 `no-candidate-accepted` 和 No-Candidate Reason；旧任务可记录
 `checked-none: <一句话原因>`。
 
 ## Harness Ledger 回写

package/scripts/check-harness.mjs

@@ -63,6 +63,8 @@ const allowedWalkthroughSkip =
   /walkthrough skipped-with-reason:\s*(docs-only|no-runtime|superseded|historical-backfill|owner-deferred)/i;
 const lessonsCreatedPattern = /checked-created:\s*(L-\d{4}-\d{2}-\d{2}-\d{3}|L-\d+)/i;
 const lessonsNonePattern = /checked-none:\s*\S+/i;
+const lessonsCandidatePattern = /checked-candidate:\s*(LC-[A-Za-z0-9-]+)/i;
+const lessonsQueuedPromotionPattern = /queued-promotion:\s*(LC-[A-Za-z0-9-]+)/i;
 
 const failures = [];
 const warnings = [];
@@ -387,15 +389,15 @@ function checkGovernanceContent() {
   if (!exists(governancePath)) return;
   const content = read(governancePath);
   const requiredTerms = [
-    "Repo Platform Profile",
-    "Branch Model",
-    "PR Policy",
-    "Required Checks",
-    "Branch Protection",
-    "Worktree Concurrency",
+    ["Repo Platform Profile", "仓库平台画像"],
+    ["Branch Model", "分支模型"],
+    ["PR Policy", "PR 规则"],
+    ["Required Checks", "必要检查"],
+    ["Branch Protection", "分支保护"],
+    ["Worktree Concurrency", "Worktree 并发"],
   ];
-  for (const term of requiredTerms) {
-    if (!content.includes(term)) fail(`${governancePath} missing section: ${term}`);
+  for (const terms of requiredTerms) {
+    if (!contentIncludesAny(content, terms)) fail(`${governancePath} missing section: ${terms[0]}`);
   }
   if (!statusWords.some((status) => content.includes(status))) {
     fail(`${governancePath} does not use evidence status model`);
@@ -408,13 +410,13 @@ function checkCiCdContent() {
   if (!exists(ciPath)) return;
   const content = read(ciPath);
   const requiredTerms = [
-    "CI Profile",
-    "Workflow",
-    "Required Checks",
-    "Evidence Status",
+    ["CI Profile", "CI 画像"],
+    ["Workflow", "工作流"],
+    ["Required Checks", "必要检查"],
+    ["Evidence Status", "证据状态"],
   ];
-  for (const term of requiredTerms) {
-    if (!content.includes(term)) fail(`${ciPath} missing section: ${term}`);
+  for (const terms of requiredTerms) {
+    if (!contentIncludesAny(content, terms)) fail(`${ciPath} missing section: ${terms[0]}`);
   }
   if (!statusWords.some((status) => content.includes(status))) {
     fail(`${ciPath} does not use evidence status model`);
@@ -428,14 +430,14 @@ function checkDeliveryOperatingModelContent() {
   const content = read(deliveryPath);
   const normalized = content.toLowerCase();
   const requiredTerms = [
-    "operating model profile",
-    "work decomposition rule",
-    "agent visibility",
-    "integration owner",
-    "delivery ssot",
+    ["operating model profile", "运行模型"],
+    ["work decomposition rule", "工作拆分规则"],
+    ["agent visibility", "agent 可见性"],
+    ["integration owner", "集成 owner"],
+    ["delivery ssot", "交付 ssot"],
   ];
-  for (const term of requiredTerms) {
-    if (!normalized.includes(term)) fail(`${deliveryPath} missing section: ${term}`);
+  for (const terms of requiredTerms) {
+    if (!contentIncludesAny(normalized, terms.map((term) => term.toLowerCase()))) fail(`${deliveryPath} missing section: ${terms[0]}`);
   }
   if (!/solo-orchestrator|team-feature-lead|split-repo-contract|program-multi-repo|waterfall-stage-gate|kanban-continuous/.test(content)) {
     fail(`${deliveryPath} does not define a recognized operating model`);
@@ -480,7 +482,7 @@ function checkReviewTemplate() {
   const reviewPath = "docs/09-PLANNING/TASKS/_task-template/review.md";
   if (!exists(reviewPath)) return;
   const content = read(reviewPath);
-  if (!content.includes("Confidence Challenge")) {
+  if (!contentIncludesAny(content, ["Confidence Challenge", "信心挑战"])) {
     fail(`${reviewPath} missing Confidence Challenge`);
   }
   if (/\|\s*R-001\s*\|\s*P[01]\s*\|.*\|\s*open\s*\|/i.test(content)) {
@@ -519,6 +521,14 @@ function columnIndex(header, pattern) {
   return header.findIndex((cell) => pattern.test(cell));
 }
 
+function contentIncludesAny(content, terms) {
+  return terms.some((term) => (term instanceof RegExp ? term.test(content) : content.includes(term)));
+}
+
+function columnIndexAny(header, patterns) {
+  return header.findIndex((cell) => patterns.some((pattern) => pattern.test(cell)));
+}
+
 function checkDuplicateIds(rows, sourcePath) {
   const seen = new Set();
   for (const cells of rows) {
@@ -535,9 +545,9 @@ function checkCloseoutSsot() {
   if (!exists(closeoutPath)) return;
 
   const closeoutContent = read(closeoutPath);
-  for (const term of ["Walkthrough", "Lessons Check", "Closeout Status"]) {
-    if (!closeoutContent.includes(term)) {
-      fail(`${closeoutPath} missing required closeout column or section: ${term}`);
+  for (const terms of [["Walkthrough"], ["Lessons Check", "Lessons 检查"], ["Closeout Status", "收口状态"]]) {
+    if (!contentIncludesAny(closeoutContent, terms)) {
+      fail(`${closeoutPath} missing required closeout column or section: ${terms[0]}`);
     }
   }
   checkNoGenericPlaceholders(closeoutPath);
@@ -545,7 +555,7 @@ function checkCloseoutSsot() {
   const closeoutTable = markdownTable(closeoutContent);
   const closeoutHeaderIndex = findHeaderIndex(closeoutTable, /^Harness ID$/i);
   const closeoutHeader = closeoutHeaderIndex >= 0 ? closeoutTable[closeoutHeaderIndex] : [];
-  const lessonsColumn = columnIndex(closeoutHeader, /^Lessons Check$/i);
+  const lessonsColumn = columnIndexAny(closeoutHeader, [/^Lessons Check$/i, /^Lessons 检查$/i, /^Lessons 检查\s*\/\s*Lessons Check$/i]);
   if (lessonsColumn < 0) {
     fail(`${closeoutPath} missing Lessons Check column`);
   }
@@ -553,10 +563,11 @@ function checkCloseoutSsot() {
   if (!exists("docs/Harness-Ledger.md")) return;
   const ledgerContent = read("docs/Harness-Ledger.md");
   const lessonIds = collectLessonIds();
+  const lessonCandidateIds = collectLessonCandidateIds();
   const ledgerTable = markdownTable(ledgerContent);
   const ledgerHeaderIndex = findHeaderIndex(ledgerTable, /^ID$/i);
   const ledgerHeader = ledgerHeaderIndex >= 0 ? ledgerTable[ledgerHeaderIndex] : [];
-  const ledgerLessonsColumn = columnIndex(ledgerHeader, /^Lessons Check$/i);
+  const ledgerLessonsColumn = columnIndexAny(ledgerHeader, [/^Lessons Check$/i, /^Lessons 检查$/i, /^Lessons 检查\s*\/\s*Lessons Check$/i]);
   if (ledgerLessonsColumn < 0) {
     fail("docs/Harness-Ledger.md missing Lessons Check column");
   }
@@ -587,20 +598,26 @@ function checkCloseoutSsot() {
     if (lessonsColumn >= 0) {
       const lessonsCheck = closeout[lessonsColumn] || "";
       const createdMatch = lessonsCheck.match(lessonsCreatedPattern);
-      if (!createdMatch && !lessonsNonePattern.test(lessonsCheck)) {
-        fail(`${closeoutPath} row ${id} needs Lessons Check value: checked-created:<lesson-id> or checked-none:<reason>`);
+      const candidateMatch = lessonsCheck.match(lessonsCandidatePattern) || lessonsCheck.match(lessonsQueuedPromotionPattern);
+      if (!createdMatch && !candidateMatch && !lessonsNonePattern.test(lessonsCheck)) {
+        fail(`${closeoutPath} row ${id} needs Lessons Check value: checked-created:<lesson-id>, checked-candidate:<candidate-id>, queued-promotion:<candidate-id>, or checked-none:<reason>`);
       } else if (createdMatch && !lessonIds.has(createdMatch[1])) {
         fail(`${closeoutPath} row ${id} references missing Lessons SSoT id: ${createdMatch[1]}`);
+      } else if (candidateMatch && !lessonCandidateIds.has(candidateMatch[1])) {
+        fail(`${closeoutPath} row ${id} references missing lesson candidate id: ${candidateMatch[1]}`);
       }
     }
 
     if (ledgerLessonsColumn >= 0) {
       const ledgerLessonsCheck = cells[ledgerLessonsColumn] || "";
       const ledgerCreatedMatch = ledgerLessonsCheck.match(lessonsCreatedPattern);
-      if (!ledgerCreatedMatch && !lessonsNonePattern.test(ledgerLessonsCheck)) {
-        fail(`docs/Harness-Ledger.md row ${id} needs Lessons Check value: checked-created:<lesson-id> or checked-none:<reason>`);
+      const ledgerCandidateMatch = ledgerLessonsCheck.match(lessonsCandidatePattern) || ledgerLessonsCheck.match(lessonsQueuedPromotionPattern);
+      if (!ledgerCreatedMatch && !ledgerCandidateMatch && !lessonsNonePattern.test(ledgerLessonsCheck)) {
+        fail(`docs/Harness-Ledger.md row ${id} needs Lessons Check value: checked-created:<lesson-id>, checked-candidate:<candidate-id>, queued-promotion:<candidate-id>, or checked-none:<reason>`);
       } else if (ledgerCreatedMatch && !lessonIds.has(ledgerCreatedMatch[1])) {
         fail(`docs/Harness-Ledger.md row ${id} references missing Lessons SSoT id: ${ledgerCreatedMatch[1]}`);
+      } else if (ledgerCandidateMatch && !lessonCandidateIds.has(ledgerCandidateMatch[1])) {
+        fail(`docs/Harness-Ledger.md row ${id} references missing lesson candidate id: ${ledgerCandidateMatch[1]}`);
       }
     }
   }
@@ -621,21 +638,45 @@ function collectLessonIds() {
   );
 }
 
+function collectLessonCandidateIds() {
+  const root = filePath("docs/09-PLANNING");
+  const ids = new Set();
+  if (!fs.existsSync(root)) return ids;
+  function visit(dir) {
+    for (const entry of fs.readdirSync(dir)) {
+      const full = path.join(dir, entry);
+      const stat = fs.statSync(full);
+      if (stat.isDirectory()) {
+        if (entry === "_archive") continue;
+        visit(full);
+        continue;
+      }
+      if (entry !== "lesson_candidates.md") continue;
+      const content = fs.readFileSync(full, "utf8");
+      for (const match of content.matchAll(/\|\s*(LC-[A-Za-z0-9-]+)\s*\|/g)) {
+        ids.add(match[1]);
+      }
+    }
+  }
+  visit(root);
+  return ids;
+}
+
 function checkLessonsSsot() {
   const lessonsPath = "docs/01-GOVERNANCE/Lessons-SSoT.md";
   if (!exists(lessonsPath)) return;
 
   const content = read(lessonsPath);
-  if (!/Detail Doc/i.test(content)) {
+  if (!contentIncludesAny(content, [/Detail Doc/i, "详情文档"])) {
     fail(`${lessonsPath} missing Detail Doc column`);
   }
 
   const table = markdownTable(content);
   const headerIndex = findHeaderIndex(table, /^ID$/i);
   const header = headerIndex >= 0 ? table[headerIndex] : [];
-  const detailColumn = columnIndex(header, /^(Detail Doc|Detail)$/i);
+  const detailColumn = columnIndexAny(header, [/^(Detail Doc|Detail)$/i, /^详情文档$/i, /^详情文档\s*\/\s*Detail Doc$/i]);
   const idColumn = columnIndex(header, /^ID$/i);
-  const statusColumn = columnIndex(header, /^Status$/i);
+  const statusColumn = columnIndexAny(header, [/^Status$/i, /^状态$/i, /^状态\s*\/\s*Status$/i]);
   if (idColumn < 0) fail(`${lessonsPath} missing ID column`);
   if (detailColumn < 0) fail(`${lessonsPath} missing Detail Doc column`);
   if (statusColumn < 0) fail(`${lessonsPath} missing Status column`);
@@ -675,7 +716,7 @@ function checkWalkthroughTemplate() {
   const walkthroughTemplate = "docs/10-WALKTHROUGH/_walkthrough-template.md";
   if (!exists(walkthroughTemplate)) return;
   const content = read(walkthroughTemplate);
-  if (!content.includes("Lessons Reflection")) {
+  if (!contentIncludesAny(content, ["Lessons Reflection", "Lessons 回看"])) {
     fail(`${walkthroughTemplate} missing Lessons Reflection section`);
   }
 }