wlvh · wlvh · Aug 7, 2026 · Aug 6, 2026 · Aug 7, 2026 · Aug 7, 2026
diff --git a/AGENTS.md b/AGENTS.md
@@ -2,6 +2,8 @@
 
 本仓库发布双语下游工作流模板、canonical `workflow-docs-sync` Skill 和安装器。
 
+新增 marker、alias、parser 或其他控制机制前，先按 DEC-006 证明独立风险、真实消费者和可复现失败路径；否则保持零 diff。
+
 - `zh/` 是中文语义源；`en/` 是从中文派生的模板与说明。
 - 根 `.github/` 是本仓库 CI 与 GitHub 基础设施；`zh/.github/`、`en/.github/` 是下游模板源。
 - 修改模板：先改 `zh/` 对应九份核心文件，再同步 `en/`。
@@ -11,16 +13,6 @@
   `zh/docs/development_workflow/`，实现决策在其 `decisions.md`。
 - 修改测试：进入 `tests/test_workflow_docs_sync.py`；具体约束以测试代码为准，不在此复制。
 
-新增 marker、alias、机器状态、parser、兼容入口或其他控制机制前，必须先证明它表达独立状态，
-并存在真实消费者与可复现失败路径。若扩展现有机制可以覆盖：
-
-- 先登记 finding；
-- 说明最小扩展边界与新增成本；
-- 不得改用同义 alias；
-- 若与用户指定实现不同，先说明替代方案。
-
-缺少真实缺口或消费者证据时保持零 diff，不因用户给出了具体实现方案就机械执行。
-
 最短验证入口：
 
 ```bash

diff --git a/en/.github/pull_request_template.md b/en/.github/pull_request_template.md
@@ -18,7 +18,7 @@ plans, historical drafts, or local uncommitted content.
 
 ## 4. Documentation Impact
 
-<!-- List documents actually updated and their evidence. Give a real no-update reason for affected candidate documents left unchanged; not every document needs a diff. -->
+<!-- List documents actually updated and their evidence. State the current factual basis for affected authorities left unchanged. -->
 
 ## 5. User-visible and Architecture Impact
 
@@ -34,7 +34,7 @@ plans, historical drafts, or local uncommitted content.
 
 ## 7. Review / Fix Record
 
-<!-- For each finding, record a stable ID, severity, first-seen round or source, evidence, judgment, fix, and recheck result. Finding states are OPEN / CLOSED / DEFERRED. REOPENED is an event, not a state; it preserves the original ID and appends evidence. CURRENT / SUPERSEDED describes only a candidate or evidence, never finding state or an overwritten earlier failure. A prior miss explanation must be evidence-backed, labeled hypothesis, or written as unknown. Formal raw records may live in comments while the body keeps an index. Record Promoted reusable rule: <authority / None>. If there are no findings, list the high-risk areas checked. Do not add a duplicate reconciliation ledger. -->
+<!-- Record the actual review, actionable feedback, disposition and recheck results, and any open decisions or limits according to this project's delivery policy. -->
 
 ## 8. Known Limits, Open Decisions, and Rollback
 
@@ -45,6 +45,5 @@ plans, historical drafts, or local uncommitted content.
 - [ ] Actual Change Scope matches the real diff.
 - [ ] Test commands, scope, results, and not-run reasons are accurate.
 - [ ] User-visible and architecture impact were checked against their authoritative documents.
-- [ ] BLOCKERs and actionable WARNs are closed; open decisions are not presented as complete.
-- [ ] The PR body contains no historical draft, unimplemented plan, wrong base/head, or repository-local
-  temporary body path.
+- [ ] Actionable review feedback required by this project is handled or explicitly recorded as open.
+- [ ] The PR body contains no historical draft, unimplemented plan, or wrong base/head.
diff --git a/en/AGENTS.md b/en/AGENTS.md
@@ -43,8 +43,8 @@ Describe stable modules, entrypoints, and responsibilities. Do not permanently m
   and the business guide. For user-visible behavior changes, update or confirm `interact.md` first.
 - Tests remain factual evidence; keep exact commands, fixtures, layers, and isolation details in
   `TESTING.md` only.
-- Not every change requires every document to change. Give a current, evidence-based no-update reason for
-  each affected candidate document left unchanged.
+- Not every change requires every document to change. For an affected authority left unchanged, state the
+  current factual basis in the delivery record.
 - Derive encoding, lint, formatter, build, and type rules from real repository configuration, not this
   template.
 
@@ -78,8 +78,8 @@ verified before execution.
 
 ## SOP
 
-Read the corresponding `SOP.md` entry for a standard process. Keep execution checklists in the current
-session; do not create repository run state, receipts, or temporary process documents.
+Read the corresponding `SOP.md` entry for a standard process. Follow this project's actual audit,
+recoverability, and delivery policy for whether execution records are stored, where, and for how long.
 
 ## PR Delivery
 

diff --git a/en/PR_Checklist.md b/en/PR_Checklist.md
@@ -31,8 +31,8 @@ explicitly requests it.
 ## Documentation and Contracts
 
 - [ ] Check `AGENTS.md`, `architecture.md`, `capability_contract.json`, `interact.md`, the business guide,
-  `TESTING.md`, and `SOP.md` according to actual impact. Give a real no-update reason for affected candidates
-  left unchanged; do not edit every document merely for completeness.
+  `TESTING.md`, and `SOP.md` according to actual impact. State the current factual basis for affected
+  authorities left unchanged; do not edit every document merely for completeness.
 - [ ] Keep the authority direction `capability_contract.json → interact.md → business_user_guide.md` for
   capability changes. User-visible claims have current implementation or test evidence and stable anchors.
 - [ ] Reference Markdown anchors through the capability contract's contract-defined protocol without copying
@@ -45,14 +45,8 @@ explicitly requests it.
 
 - [ ] Complete the review gate required by this project's test and delivery policy, and accurately record
   reviewer identity, scope, and limitations.
-- [ ] Fix every BLOCKER and actionable WARN that does not require a new product decision. Keep remaining
-  issues in open decisions with evidence and impact.
-- [ ] Preserve material rounds, REOPENED findings, and SUPERSEDED candidates or evidence. Reopening keeps the
-  original finding ID, and finding state is not confused with evidence state.
-- [ ] Support a prior miss explanation with evidence or label it `hypothesis` / `unknown`; never invent cause
-  merely to fill a field.
-- [ ] Evaluate every material finding for promotion to a long-term rule or automated gate, and record why when
-  it is not promoted.
+- [ ] Handle actionable review feedback according to this project's policy. Keep recheck evidence for resolved
+  items and record the impact and required decision for anything left open.
 - [ ] Rerun affected tests and mechanical checks after fixes, then recheck the final diff and Git state.
 
 ## PR Delivery

diff --git a/en/README.md b/en/README.md
@@ -28,6 +28,10 @@ Review prefers a fresh-context, blind-first independent reviewer. When cognitive
 the result is reported honestly as self-review. The deterministic checker proves final repository state
 only, not investigation, test, or review history.
 
+The installer copies only the canonical Skill. Before any target mutation it rejects symlinks, frontmatter
+without standard delimiters, and ignored source residue that would otherwise be copied; it stores no source
+receipt.
+
 ## Template Contract
 
 Markdown project-fill slots use `<!-- project-fill: ... -->`; JSON uses strings prefixed with

diff --git a/en/SOP.md b/en/SOP.md
@@ -4,8 +4,8 @@
 
 `SOP.md` keeps stable process entrypoints and does not copy volatile commands, test lists, or publishing
 details. When sources conflict, current code, configuration, tests, contracts, and focused authorities such
-as `TESTING.md` and `PR_Checklist.md` take precedence. Keep execution checklists in the current session;
-do not write them as repository run state, receipts, or temporary SOPs.
+as `TESTING.md` and `PR_Checklist.md` take precedence. Follow this project's actual audit, recoverability,
+and delivery policy for how execution records are stored and retained.
 
 ## Available SOPs
 

diff --git a/en/TESTING.md b/en/TESTING.md
@@ -45,9 +45,9 @@ maintenance value.
 
 ## 3. Capability Contract Alignment
 
-An alignment test belongs to the target project's test suite, not the documentation sync checker. It
-should recursively collect stable `anchor_id` values from every object in `capability_contract.json` and
-use the protocol defined by the contract rules to check uniqueness and Markdown references without
+A project's alignment test should recursively collect stable `anchor_id` values from every object in
+`capability_contract.json`. It should use the protocol defined by the contract rules to check uniqueness
+and Markdown references without
 hardcoding buckets, JSON paths, array positions, or requiring every contract entry to appear in the
 business guide.
 

diff --git a/en/docs/development_workflow/README.md b/en/docs/development_workflow/README.md
@@ -68,7 +68,14 @@ external shallow clone, then pins target HEAD and upstream SHA for the entire ru
   source and validates final HEAD, dirty scope, no index/worktree split on editable paths, nine regular UTF-8
   nonempty files, a JSON object, active markers, and an existing `.gitignore` as UTF-8. Final-byte whitespace
   runs with fixed Git rules from a temporary non-repository directory, independent of target attributes. It
-  does not parse Markdown or validate capability truth, test levels, prose quality, or execution history.
+  does not parse Markdown or validate capability truth, test levels, prose quality, or execution history. The
+  single final-byte path depends on rejecting every dirty path outside the allowlist; relaxing that allowlist
+  requires reevaluating whitespace coverage. Split detection aggregates both status sides by path and covers
+  an index deletion or rename source followed by an untracked or ignored recreation at that path. It removes
+  two publication candidates rather than running a second whitespace check.
+- Before any removal or copy, the installer validates source and destination ancestors, source symlinks,
+  standard delimiters around Claude frontmatter, and ignored source residue that would be copied. It only
+  removes the exact obsolete reviewer Skill and stores no install state or source receipt.
 - Keep temporary PR body Markdown outside the repository. Commit, push, and draft-PR creation occur through
   general GitHub publishing capability only after successful checks and only when the user asks.
 
@@ -86,13 +93,18 @@ not promise exhaustive alias detection, and structural references do not prove c
 does not raise `schema_version` from `0.1.0`. TESTING defines minimum evidence for escaped bugs, public
 contracts, no-test diffs, refactors, and documentation-only gates, while the checklist enforces the
 add-or-not-add test decision.
-Findings keep stable IDs and first-seen evidence; REOPENED is an event, CURRENT / SUPERSEDED applies only to
-candidates or evidence, and prior misses are evidence-backed, labeled hypothesis, or recorded as unknown.
-
-With the same target base, upstream candidate, language, and round-one document bytes, Case A round two passes
-only as `PASS_NOOP`; an added valid correction is `ROUND1_INCOMPLETE`, and unsupported prose drift is
-`ROUND2_DRIFT`. Either failure restarts both rounds from a clean target. This decision adds no parser, ledger,
-receipt, run state, installer behavior, or `sync_docs.py` / CLI feature.
+Stable IDs, first-seen evidence, REOPENED events, and candidate/evidence states belong to this repository's
+maintenance and canonical Skill/eval evidence contract. Downstream templates defer review records,
+actionable feedback, rechecks, and open decisions to target-project policy instead of fixing this vocabulary.
+
+Case A preregisters the selected target and known-stale claims before starting a blind executor. Target code,
+configuration, tests, other committed artifacts, upstream candidate, and language stay fixed; round one
+commits only its final nine documents, and round two starts from that `second_target_sha` in a new clean
+checkout. A zero document diff with no staged, untracked, or ignored residue is `PASS_NOOP`; an added valid
+correction is `ROUND1_INCOMPLETE`, and unsupported prose drift is `ROUND2_DRIFT`. Alignment runs directly on
+the committed second target instead of a derived test-only identity. Either failure restarts both rounds from
+a clean target. This decision adds no parser, ledger, receipt, run state, installer behavior, or
+`sync_docs.py` / CLI feature.
 
 Key implementation files: