@geoly-ai/social-hub-cli 0.3.2 → 0.3.4

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@geoly-ai/social-hub-cli",
-  "version": "0.3.2",
+  "version": "0.3.4",
   "type": "module",
   "description": "social-hub CLI for Social Ops Hub",
   "repository": {
@@ -22,8 +22,8 @@
   },
   "dependencies": {
     "commander": "^12.1.0",
-    "@geoly-ai/social-hub-authz": "0.0.8",
-    "@geoly-ai/social-hub-sdk": "0.0.44"
+    "@geoly-ai/social-hub-authz": "0.0.9",
+    "@geoly-ai/social-hub-sdk": "0.0.45"
   },
   "devDependencies": {
     "@types/node": "^22.10.2",

package/skills/manifest.json

@@ -1,7 +1,7 @@
 {
   "version": 1,
   "package": "social-ops-hub",
-  "cliVersion": "0.3.2",
+  "cliVersion": "0.3.4",
   "repository": "social-ops-hub",
   "publish": {
     "registry": "https://skill.sh",

package/skills/reddit-content-rewrite/SKILL.md

@@ -135,8 +135,9 @@ description: >-
 HandoffRepair · ArtifactType=Post（受管项目主帖）必需：
   OriginalContent / RequestedRepair / ContentKey /
   ContentFunction / PrimaryPostType / StructureFamily /
+  AssignedFlair / PostingTrigger /
   OpeningMove / EvidenceShape / BrandEntryMode / CaveatMode /
-  EndingMode / LengthBand / AnchorSlot / SecondaryAnchorSlot / SampleSetID /
+  EndingMode / LengthBand / AnchorSlot / SecondaryAnchorSlot / SampleSetID / FlairCoverageStatus /
   （SecondaryAnchorSlot 仅用于 >350 词且有两个 ABU 的主帖，否则填 `—`）
   AllowedStructuralChanges / ForbiddenStructuralChanges
   （B 层要"保持全部指纹字段"，就必须收到全部原始值；缺任一项 → BlockingGap，
@@ -405,6 +406,48 @@ ExplicitStructureSpec:
 完整命令载荷、失败处理和记忆记录格式见
 `../reddit-strategy-shared/references/content-review-gate.md`。当前 Hub 服务最多支持两次尝试；第二次仍不通过时不得无限重试、降低标准或把结果标成通过。
 
+**Step 3.6 结构根因诊断（`StructuralRootCause`）—— 别在措辞层无限空转。**
+
+措辞层（`AllowedTextualRepairs`）连续两轮修复后，若 `ai_structural_tone` / 立意 /
+叙事前提仍不达标，就不该继续在措辞层打转——问题的根在结构 / Brief，须退回 owner。
+
+**判定分工（四字段互不重叠）：**
+
+| 字段 | 回答 |
+|---|---|
+| `StructureChangeRequired` | 当前要求是否明确需要改动受保护结构面（锁定字段 / outline premise / paragraph order）？ |
+| `StructuralChangeStatus` | rewrite 实际是否越权改了结构？ |
+| `HubStructureConflictStatus` | Hub 的 `optimized` 是否触及受保护字段？ |
+| `StructuralRootCause` | 结构保持型措辞修复耗尽后，剩余失败是否根源于 Brief / 立意 / 结构？ |
+
+**`StructuralRootCause: Yes` 必须同时满足全部六条**（缺一即不得判 Yes）：
+
+1. 已完成规定的结构保持型 repair 流程（Round 1 = 初稿本地 self-check + textual repair → Hub attempt 1；Round 2 = 按 attempt 1 issues 做结构保持型修复 → Hub attempt 2）；
+2. Hub attempt 2 仍非 `pass`；
+3. 剩余阻断问题与第一次属**同一结构性问题簇**（`ai_structural_tone` / 立意 / 叙事前提 / 开场 / 品牌进入 / 结尾 / 信息拓扑）；
+4. 已如实记录两轮具体 `AllowedTextualRepairs`，不是只写「已去 AI 味」；
+5. 继续解决将**不可避免地**触及 `LockedFingerprintField` / outline premise / paragraph order；
+6. 不存在更优先的**事实 / 合规 / 证据 / 上下文 / 账号绑定 / Hub 不可用**缺口（这些各有其阻断路径，不得混记为结构根因）。
+
+> **不得把「第二次不 pass」自动等同于 `StructuralRootCause: Yes`。** attempt 2 仍失败
+> 也可能是证据不足、披露缺失、subreddit 上下文缺失、措辞修复没做对、或 Hub degraded。
+
+触发后输出（退回结构 owner，不在本 skill 自行改结构）：
+
+```markdown
+StructuralRootCause: Yes
+StructureChangeRequired: Yes
+AffectedLockedFields: <明确字段；无法确定则 PendingBriefDiagnosis>
+FixOwnerSkill: reddit-phase-brief
+ReadyForUse: No
+```
+
+此时 `HubStructureConflictStatus` 仍可为 `None`（Hub 未必擅自改结构，只是指出当前
+结构过不了；只有 `optimized` 实际触及受保护字段才是 `Escalated`）。证据字段：
+`TextualRepairRounds` / `TextualRepairsApplied` / `RemainingBlockingIssues` /
+`SameIssuePersistedAcrossAttempts` / `RequiredStructuralSurface` /
+`StructuralRootCauseEvidence`。
+
 **Step 4** 仅在 Hub 返回 `completed/pass` 后按内容类型选择交付格式。不要把内部提示词模板、完整检查清单或评分维度原样展示给用户；默认只交付已审核的最终稿。若本次是 ContentOnly，交付物必须标为“内容审核通过草稿”，同时保留 `AccountBindingStatus: Pending`、`ReadyForDelivery: No`；只有 AccountBound 才能称为可发布终稿。若 Hub 不可用，交付状态必须明确为未通过/不可用，而不是输出“可发布”内容。
 
 | 内容类型 | 默认交付格式 | 何时附加说明 |
@@ -433,7 +476,7 @@ ExplicitStructureSpec:
 - 不编造作者身份、国籍、语言背景
 - **不主动制造 typo、拼写失误、错误冠词、介词误用、缺主语、"移民语法"**——
   这些不只是同质化，是伪造
-- 所有 claim 必须可追溯到 `KnownProductFacts` / `SourceEvidence`；追溯不到就删 claim 或 BlockingGap
+- 所有进入可发布文本（标题 / 正文 / 评论 / 回复）的实质性 claim 必须是 `ClaimEvidenceStatus: Cited`（可映射到已批准证据 `KnownProductFacts` / `SourceEvidence`）或 `ExternallyVerified`（非产品事实类、有外部权威来源 `ExternalSourceRef`）；`Inferred`（无来源的模型推断）禁入可发布文本——删除或退回补证据，无法删且必需则 BlockingGap。三态权威定义见 `../reddit-strategy-shared/references/evidence-policy.md`
 
 ### 绝对禁用营销词
 
@@ -613,7 +656,7 @@ A / B 层每帖必查；C 层在其作用域内每帖必查（见下）。
 - [ ] 未编造缺点（不得为显得真实而插入安全 caveat）
 - [ ] 未主动制造 typo、错误冠词、介词误用、"移民语法"等伪真人信号
 - [ ] 未编造作者身份、国籍、语言背景
-- [ ] 所有 claim 可追溯到 `KnownProductFacts` / `SourceEvidence`（含 SKU 基准表、VOC、甲方语料、用户提供证据）
+- [ ] 所有进入可发布文本（标题 / 正文 / 评论 / 回复）的实质性 claim 为 `Cited` 或 `ExternallyVerified`（`ClaimEvidenceStatus` 三态，定义见 `../reddit-strategy-shared/references/evidence-policy.md`）；`Inferred`（模型推断、无来源）**禁入可发布文本**，只能留在 Brief 内部推理或删除，无法删且为必需则 `BlockingGap`。产品性能 / 亲身体验 / 使用时长不因搜到第三方页面就可声称，仍须 `Cited` 到已批准证据
 - [ ] 无禁用营销词
 - [ ] 披露、品牌风险、subreddit 合规规则已满足；已有披露未被删除或移到失效位置
 - [ ] 未写成伪求助（标题问句但正文全是推荐，或为后续品牌评论制造入口）
@@ -642,6 +685,7 @@ A / B 层每帖必查；C 层在其作用域内每帖必查（见下）。
 - [ ] `LengthBand` 未变；仅当明确允许长度变化时，才检查变更后的档位是否与 `SampleSetID` 兼容；无 baseline 时保持原文
 - [ ] `AnchorSlot` 合法且未被移动
 - [ ] `SecondaryAnchorSlot`（仅 >350 词且两个 ABU 时；否则为 `—`）合法且未被移动
+- [ ]（仅 `ArtifactType=Post`）`PostingTrigger` 保持不变：改后正文反推的 trigger 仍等于传入值，且不违反 `PostingTrigger × BrandEntryMode` 交叉约束（`UnsolicitedTestimony` ⇒ `BrandEntryMode: 不出现`）。它是 `PrewritingControlField`、**不是** `LockedFingerprintField`，因此漂移记 `PostingTriggerStatus: Drift`（不并入 `StructuralChangeStatus`）；按 handoff 规则**标记而非应用**该变化，退回 owner，不由 rewrite 自行接受
 - [ ] 未被改写成"先痛点后产品"的推荐弧光（除非本篇 family 本就是 `推荐转化叙事`）
 - [ ] 未被强行添加问题结尾
 - [ ] 未被强行添加限制说明
@@ -737,6 +781,10 @@ subreddit 规则、属于 `SampleSetID` 支持的异常值，否则保持原文
   #                                           spec 不能顶替（ManagedContentRequiresBriefStructureFamily）
   #   ZeroDraftGeneration ad-hoc (Post)     → StructureFamily 或 ExplicitStructureSpec 任一
   #   StandaloneSourceRepair                → 均可 NotProvided，靠 ObservedSource
+- AssignedFlair:                  # PrewritingControlField；Post 传入并保持；Comment/Reply/Title 填 NotApplicable
+- FlairCoverageStatus:            # Exact | ApprovedFallback | NotApplicable；Post 透传，不在 rewrite 丢失
+- PostingTrigger:                 # PrewritingControlField 受控五值；Post 保持并做 Planned-vs-final；不进碰撞三元组；Comment/Reply/Title 填 NotApplicable
+- PostingTriggerStatus: Preserved | Drift | NotApplicable   # 反推 trigger 与传入值是否一致 + 交叉约束
 - OpeningMove:
 - EvidenceShape:
 - BrandEntryMode:
@@ -754,6 +802,13 @@ subreddit 规则、属于 `SampleSetID` 支持的异常值，否则保持原文
 - ForbiddenStructuralChanges:
 - StructuralChangeStatus: Unchanged | ChangedWithinAllowance | BlockingDrift
 - ChangedFingerprintFields: None | <field list>
+- ClaimEvidenceAllCitedOrExternallyVerified: Yes | No | NotApplicable   # no Inferred claim in any publishable text (title/body/comment/reply)
+- TextualRepairRounds: <n>
+- SameIssuePersistedAcrossAttempts: Yes | No | NotApplicable
+- StructuralRootCause: Yes | No                # Yes 须满足全部六条判据
+- StructureChangeRequired: Yes | No             # 需改动受保护结构面（锁字段 / outline premise / paragraph order）；StructuralRootCause=Yes 时随之为 Yes
+- AffectedLockedFields:                         # StructuralRootCause=Yes 时填；无法确定 PendingBriefDiagnosis
+- StructuralRootCauseEvidence:                  # 两轮 TextualRepairsApplied + RemainingBlockingIssues + RequiredStructuralSurface
 
 - DiversityStatus: Pass | SoftWarning | BlockingCollision | NotApplicable
 - DiversityClaimed: Yes | No

package/skills/reddit-content-writing/SKILL.md

@@ -47,6 +47,8 @@ Additionally, stop before writing a single word of body copy if any of these hol
 
 - an execution-table row carries a `客户内容原型` code that is not mapped in `memory/phase0-客户内容原型映射.md` → `RoutingStatus: BlockingGap`, `AllowedToFallbackToGenericRecommendation: No`, `RequiredAction: Strategy mapping approval`
 - a main-post row has no `StructureFamily` assigned by the Brief
+- a main-post row has no `AssignedFlair`, or the baseline card's `FlairCoverageStatus` is `Insufficient` → `BlockingGap`, return to `reddit-phase-brief`
+- a main-post row has no `PostingTrigger` assigned by the Brief, or its `Planned` `PostingTrigger` already violates the `PostingTrigger × BrandEntryMode` cross-constraint → `BlockingGap` (the post-draft re-derive of `PostingTrigger` vs `Planned` is checked after drafting, in the Writing Rules, not here)
 - a target subreddit has no baseline card, or it is `Expired` / `Incomplete` → `BlockingGap: BaselineMissingOrStale`
 - `memory/phase4-叙事指纹索引.md` has not been read, or the rolling-window collision check has not been run
 - the collision check fails and no sample-based exception is on record → `BlockingGap: StructureCollision`, return to `reddit-phase-brief`
@@ -116,9 +118,22 @@ currently allows two attempts; if the second result is not `pass`, stop with
 
 ## Writing Rules
 
+**W0 flow (fixed order — 先取样 → 再写 → 再审 → 再改):** the whole batch runs
+`AssignedFlair → flair-level baseline → structure pre-assign (Brief) → draft →
+local A/B/C self-check → Hub review → allowed textual repair → Hub re-review →
+pass then hash`. Writing never re-samples to change structure and never treats a
+Hub `pass` as re-openable — any later body change voids the Hub pass and the
+hash (see `content-diversity-contract.md`). Do **not** describe this as "Hub pass,
+then hand to rewrite to keep editing".
+
 **Structure and anti-homogenisation (v2.12.25, checked before any other rule):**
 
-- **Verify the subreddit baseline card first.** It must exist, be within its freshness window, and cover all 10 dimensions (word-count median + IQR, paragraph-count median, title-form distribution, list/subheading usage rate, first-brand-position distribution, question-ending rate, TL;DR prevalence, first-person rate, review-vs-question length split, dash/parenthesis usage). Missing or stale → `BlockingGap: BaselineMissingOrStale`, return to `reddit-phase-brief`. **Do not rebuild the baseline here.**
+- **Verify `AssignedFlair` and flair coverage first.** Every main-post Key must
+  carry a non-empty `AssignedFlair` matching the baseline card's `BaselineFlair`
+  (or an approved fallback), and the card's `FlairCoverageStatus` must be `Exact`
+  or `ApprovedFallback`. Missing flair or `Insufficient` coverage → `BlockingGap`,
+  return to `reddit-phase-brief`. Writing never invents or reassigns the flair.
+- **Verify the subreddit baseline card next.** It must exist, be within its freshness window, and cover all 10 dimensions (word-count median + IQR, paragraph-count median, title-form distribution, list/subheading usage rate, first-brand-position distribution, question-ending rate, TL;DR prevalence, first-person rate, review-vs-question length split, dash/parenthesis usage). Missing or stale → `BlockingGap: BaselineMissingOrStale`, return to `reddit-phase-brief`. **Do not rebuild the baseline here.**
 - **Read `memory/phase4-叙事指纹索引.md` and run the rolling-window collision check before writing.** The window is defined by `content-diversity-contract.md` and locked at project start: same subreddit, `WindowCount` (8–12) **and** `WindowDays` (30–60) **both in force** — not either/or. Rows count only when `RecordStatus` is `Final`/`Reviewed`/`Planned`; `Superseded`/`Abandoned` are excluded, and the current Key's own row is excluded from comparison.
 - **Hard prohibition:** the same `StructureFamily + OpeningMove + EndingMode` triple as **any other Key's row anywhere in the window** — not merely the immediately preceding post. "Adjacent posts must not match" is a strict subset of this rule and is not sufficient: if only neighbours were compared, one intervening post would license repeating the same structure forever.
 - **Collision repair order is fixed** (see below); the writer never resolves a collision by silently choosing a different family.
@@ -128,6 +143,15 @@ currently allows two attempts; if the second result is not `pass`, stop with
 - **Real limitations:** state genuine, decision-relevant limitations when they exist; never invent a safe flaw to look authentic; position and form follow the content, not a fixed paragraph.
 - **Punctuation and format values follow the subreddit baseline** — no cross-subreddit em-dash ban, no parenthesis quota, no fixed word count.
 - **ABU (Answer-Bearing Unit, B-group):** 1–2 per post, all 8 hard requirements met, `AnchorSlot` from the 7-value enum, visibility gate satisfied per length band. The old "must sit in paragraph 2 or 3" rule is removed.
+- **`PostingTrigger` is Brief-pre-assigned, not writer-chosen.** After drafting,
+  re-derive the actual `PostingTrigger` from the finished body and confirm it
+  matches the `Planned` value; a mismatch is a `BlockingGap` (return to Brief),
+  handled exactly like a locked-field drift. Enforce the cross-constraint both
+  ways: `BrandEntryMode ≠ 不出现` forbids `UnsolicitedTestimony`, and
+  `PostingTrigger = UnsolicitedTestimony` requires `BrandEntryMode: 不出现` plus a
+  real, non-solicited, non-brand-controlled testimony in `SourceEvidence` (never
+  impersonate its original author). `PostingTrigger` is recorded but is **not**
+  part of the hard collision triple.
 
 **Handoff to rewrite (structure-preserving):**
 
@@ -143,11 +167,20 @@ currently allows two attempts; if the second result is not `pass`, stop with
 - Preserve required edits from `reddit-brand-risk-response`, especially delayed brand placement, removal of CTA/link-first phrasing, balanced limitations, substantiated claims, and less generic bot-like cadence.
 - Keep a post and all paired comments together.
 - Enforce P-B/P-NB and C-B/C-NB complement rules.
-- Do not invent product facts or claims.
+- Do not invent product facts or claims. **Every substantive claim in any
+  publishable text (title / body / comment / reply) must be `Cited` or
+  `ExternallyVerified` per `ClaimEvidenceStatus`** (three-state, defined
+  in `../reddit-strategy-shared/references/evidence-policy.md`). `Inferred` claims
+  (model synthesis/inference with no source) are **forbidden from all publishable
+  text** — keep them in
+  the Brief's internal reasoning or drop them; if one cannot be removed and is
+  required to complete the piece → `BlockingGap`. Product performance / lived
+  experience / usage duration is not made claimable just because a third-party
+  page exists — it still needs `Cited` to approved evidence.
 - Do not write for subreddits that are only in `NeedsVerificationCandidates`.
 - Ask for the Feishu draft document/folder link before creating or writing documents.
 - Update `memory/phase4-撰写进度.md` after each completed batch.
-- Record every main post in `memory/phase4-叙事指纹索引.md` (fixed table: Key / Subreddit / ContentFunction / StructureFamily / OpeningMove / EvidenceShape / BrandEntryMode / CaveatMode / EndingMode / LengthBand / AnchorSlot / SampleSetID / DraftHash).
+- Record every main post in `memory/phase4-叙事指纹索引.md` (fixed table: Key / Subreddit / ContentFunction / StructureFamily / OpeningMove / EvidenceShape / BrandEntryMode / CaveatMode / EndingMode / LengthBand / AnchorSlot / SampleSetID / DraftHash), plus the required extension columns including the two `PrewritingControlFields` `AssignedFlair` and `PostingTrigger`.
 
 ## Outputs
 
@@ -168,13 +201,18 @@ currently allows two attempts; if the second result is not `pass`, stop with
 - DraftKeys:
 - PrototypeRoutingStatus: Mapped / BlockingGap
 - BaselineStatus: Fresh / Expired / Missing / Incomplete
+- FlairCoverageStatus: Exact / ApprovedFallback / Insufficient   # batch-level = worst status across per-Key cards (Insufficient > ApprovedFallback > Exact); per-Key values live in PerKeyStructureContract
+- AssignedFlairVerified: Yes / No
+- PostingTriggerPlannedVsFinal: Match / Drift / NotAssigned
+- PostingTriggerCrossConstraintOK: Yes / No
+- ClaimEvidenceAllCitedOrExternallyVerified: Yes / No   # no Inferred claim reached any publishable text (title/body/comment/reply)
 - FingerprintIndexUpdated: Yes / No
 - CollisionCheckPassed: Yes / No / ExceptionRecorded
 - PerKeyStructureContract: |
     <one row per DraftKey — batch-level single values are not valid, a batch normally
      spans several families, anchor slots and hashes>
-    | Key | ClientContentPrototype | ContentFunction | PrimaryPostType | StructureFamily | SampleSetID | AnchorSlot | SecondaryAnchorSlot | AllowedStructuralChanges | ForbiddenStructuralChanges | DraftHash |
-    |---|---|---|---|---|---|---|---|---|---|---|
+    | Key | ClientContentPrototype | ContentFunction | PrimaryPostType | StructureFamily | AssignedFlair | FlairCoverageStatus | PostingTrigger | SampleSetID | AnchorSlot | SecondaryAnchorSlot | AllowedStructuralChanges | ForbiddenStructuralChanges | DraftHash |
+    |---|---|---|---|---|---|---|---|---|---|---|---|---|---|
 - SKUBindingsPreserved: Yes / No
 - BrandWordStatusPreserved: Yes / No
 - ComplementRulesPassed: Yes / No

package/skills/reddit-content-writing/references/phase4-writing.md

@@ -25,13 +25,15 @@
 [ ] 本批每个目标 subreddit 均有 baseline card（Brief 阶段产出）
 [ ] baseline 状态为 `Fresh`（受控值：`Fresh` / `Expired` / `Missing` / `Incomplete`；默认新鲜期 60 天，项目可在 shared 契约中覆盖）
 [ ] baseline card 覆盖下表全部 10 个维度
+[ ] 每个主帖 Key 的 `AssignedFlair` 非空，且与 baseline 卡的 `BaselineFlair` / 覆盖口径一致
+[ ] baseline 卡 `FlairCoverageStatus` 为 `Exact` 或 `ApprovedFallback`（`Insufficient` → BlockingGap，退回 Brief）
 ```
 
 任一项不满足 → 标 `BlockingGap: BaselineMissingOrStale`，**退回 `reddit-phase-brief` 重建 baseline，不得在 writing 阶段自行补齐后继续开写**。
 
 **W-1.b baseline card 必备维度（v2.12.25 从「2–3 条 excerpt + 词数范围」升级）：**
 
-样本要求：**同 subreddit、同 flair / 同内容功能的 5–10 条近期真实帖**（少于 5 条即 `BlockingGap`，标注 `SampleSetID` 与实际条数）。
+样本要求：**同 subreddit、同 `AssignedFlair`（退而求其次同内容功能）的 5–10 条近期真实帖**（精确 flair 少于 5 条时按 shared 契约的 flair 覆盖门禁二分：flair 决定资格/披露/帖型/上下文 → `BlockingGap`；仅缺风格分布 → 批准后 `ApprovedFallback`。标注 `SampleSetID`、`BaselineFlair`、`FlairCoverageStatus` 与实际条数）。
 
 | 维度 | 记录形式 | 用途 |
 |------|---------|------|

package/skills/reddit-matrix.lock.json

@@ -7,7 +7,7 @@
       "contentSha256": "dfaf4bafae4118deb37b3857fea660e025a64203a9dcaf222fd0d12d4b4e9627"
     },
     "reddit-strategy-shared": {
-      "contentSha256": "9f959f956bc4e5a2dc735a9d3b4b915a210efc47467be173a531c82b5920c403"
+      "contentSha256": "5774f1d9d6bc93f04549dfdb606ff24c258eb900a43baf880e11bd13c7a27fd4"
     },
     "reddit-phase-research": {
       "contentSha256": "f1c2260587b5d6a82272efd6da1289a8bf667a05eb65fa8a774059348d2afca9"
@@ -16,13 +16,13 @@
       "contentSha256": "04cbfddebd8cb6d22de6eb9711d9fd0580680190ca8465ca7a888f44338db4c4"
     },
     "reddit-phase-brief": {
-      "contentSha256": "59c8fbdf642b2d91a6ae7f285d241837dda56c3cf0800aa40c7736707b45df41"
+      "contentSha256": "a56b9648acb48260ec06d0304d5bcc04b13e971471ba1c3a910147c472125882"
     },
     "reddit-content-writing": {
-      "contentSha256": "a54a6130e87b53d5a3bc7e1b84e3d545c39f1fb08a1caa9fa159af1bad935fd2"
+      "contentSha256": "c4aeb42db6343475428215f90ee0ae01fe2f1837f17754025b733e55c90a98bb"
     },
     "reddit-content-rewrite": {
-      "contentSha256": "f3ce42857d3d8a3cd06bbad5257a156ce42d2dcbc0d74a472c7d710bde6b0905"
+      "contentSha256": "9a2a64e08727dfaf34550f75cfa568b81ea285a030e770c54fea064e6fc4a13d"
     },
     "reddit-delivery-check": {
       "contentSha256": "e619faf2e9a54770b4ee52889bd1da045c21c5226ea8e58b6fd02b2ee87ca77b"

package/skills/reddit-phase-brief/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: reddit-phase-brief
 description: |
-  Reddit Phase 4 Brief skill for the matrix SOP. Use after reddit-phase-planning completes Phase 4.0.6 and before reddit-content-writing starts. It establishes the subreddit baseline first, then locks ContentFunction and StructureFamily, then the outline; it also creates board rationale, content schedule, per-Key title/direction/outlines, brand/SKU placement strategy, structure pre-assignment (OpeningMove / EvidenceShape / BrandEntryMode / EndingMode / LengthBand / SampleSetID / RecentCollisionCheck), and the approval gate. Trigger when the user says "确认进入 Phase 4 Brief", "开始 Brief", "产出 Brief 汇报稿", or when planning is complete but drafting has not started.
+  Reddit Phase 4 Brief skill for the matrix SOP. Use after reddit-phase-planning completes Phase 4.0.6 and before reddit-content-writing starts. It assigns the target flair first, then establishes the flair-level subreddit baseline, then locks ContentFunction and StructureFamily, then the outline; it also creates board rationale, content schedule, per-Key title/direction/outlines, brand/SKU placement strategy, prewriting control fields (AssignedFlair / PostingTrigger), structure pre-assignment (OpeningMove / EvidenceShape / BrandEntryMode / EndingMode / LengthBand / SampleSetID / RecentCollisionCheck), and the approval gate. Trigger when the user says "确认进入 Phase 4 Brief", "开始 Brief", "产出 Brief 汇报稿", or when planning is complete but drafting has not started.
 ---
 
 # Reddit Phase Brief
@@ -63,21 +63,33 @@ Read these local references before producing the brief:
 
 ## Workflow
 
-The v2.12.25 order is **baseline → ContentFunction → StructureFamily → outline
-lock**. Do not write an outline before the structure is decided from evidence.
+The v2.12.25 order is **flair → baseline → ContentFunction → StructureFamily →
+outline lock** (the W0「先取样→再写→再审→再改」重排：动笔前先确定 flair、产出 flair
+级风格卡/baseline，再决定结构、写 outline；不在写作阶段临场补采）。Do not write an
+outline before the structure is decided from evidence.
 
 1. Confirm `brief_approval_mode`: `甲方审核` or `内部确认`. If absent from `memory/hot-cache.md`, ask the user before generating the brief.
 2. Confirm the number of Brief items in this batch.
 3. Load or initialize `memory/phase4-内容排期表.md`.
 4. Check historical `memory/phase4-撰写brief.md` for overlap before creating a new batch.
-5. **Establish or refresh the subreddit baseline** for every target subreddit and assign a `SampleSetID` (see `references/phase4-brief-workflow.md` step B-1.0). Reuse a valid, unexpired baseline; only re-acquire when it is missing or stale.
-6. **Lock the project diversity window** (`WindowCount` 8–12, `WindowDays` 30–60) if not already locked, and read the rolling window from `memory/phase4-叙事指纹索引.md`.
-7. Produce board-selection rationale with evidence URLs, then the board overview table.
-8. For every post Key: resolve `客户内容原型` → `MappingRuleID`, lock `ContentFunction`, confirm the `EvidenceShape` actually in hand, narrow `StructureFamily` candidates with the ordered filter in `content-type-routing-contract.md`, pre-assign `OpeningMove` / `BrandEntryMode` / `CaveatMode` / `EndingMode` / `LengthBand` / `AnchorSlot`, and **only then** run `RecentCollisionCheck` on the complete triple. On collision, first swap `OpeningMove` / `EndingMode` within the same family; reselect a family only if that fails. **Only then** write the outline.
-9. Create one Brief block for every execution-table Key in the selected batch, including posts, self-comments, and external comments.
-10. For all P-B/C-B/PGC rows, record brand/SKU placement rationale using `references/brand-SKU-PGC-rules.md`.
-11. Write a `Planned` fingerprint row for every briefed post Key into `memory/phase4-叙事指纹索引.md` (`DraftHash: PENDING`).
-12. Write or update the memory files listed below, then show the Brief gate checklist.
+5. **Assign `AssignedFlair` (required, per main-post Key) before sampling.** flair 同时
+   决定合规资格、披露要求与触达，且决定 baseline 的采样口径。flair 必须取自该 subreddit
+   的现行 flair 列表（不得自造）；板块无 flair 或允许不选时用哨兵值 `NoFlairAvailable` /
+   `NoFlairRequired`（不得留空、不得用评论的 `N/A`）；缺失 → `BlockingGap`。评论 / 回复填 `N/A`。
+6. **Establish or refresh the flair-level subreddit baseline** for every target subreddit
+   and assign a `SampleSetID` (see `references/phase4-brief-workflow.md` step B-1.0).
+   优先按 `AssignedFlair` 细分取 exact-flair 样本；记录 `BaselineScope` /
+   `FlairCoverageStatus` / `FallbackApprovalRef`。只有 flair 仅决定风格分布、且拿到批准
+   时才可 `FlairCluster` / `SubredditFallback`；flair 决定资格 / 披露 / 帖型 / 上下文
+   却无合格样本则 `FlairCoverageStatus: Insufficient` → `BlockingGap`，不得 fallback。
+   Reuse a valid, unexpired baseline of the same flair coverage; only re-acquire when it is missing or stale.
+7. **Lock the project diversity window** (`WindowCount` 8–12, `WindowDays` 30–60) if not already locked, and read the rolling window from `memory/phase4-叙事指纹索引.md`.
+8. Produce board-selection rationale with evidence URLs, then the board overview table.
+9. For every post Key: resolve `客户内容原型` → `MappingRuleID`, lock `ContentFunction`, confirm the `EvidenceShape` actually in hand, narrow `StructureFamily` candidates with the ordered filter in `content-type-routing-contract.md`, pre-assign `OpeningMove` / `BrandEntryMode` / `CaveatMode` / `EndingMode` / `LengthBand` / `AnchorSlot`, and the `PostingTrigger`（受控值 `ConcreteFailure` / `PurchaseDecision` / `MeasurementResult` / `CompatibilityGap` / `UnsolicitedTestimony`；`UnsolicitedTestimony` 仅当 `BrandEntryMode: 不出现` 且有真实非征集非品牌控制的 testimony 证据），and **only then** run `RecentCollisionCheck` on the complete triple. On collision, first swap `OpeningMove` / `EndingMode` within the same family; reselect a family only if that fails. **Only then** write the outline.
+10. Create one Brief block for every execution-table Key in the selected batch, including posts, self-comments, and external comments.
+11. For all P-B/C-B/PGC rows, record brand/SKU placement rationale using `references/brand-SKU-PGC-rules.md`.
+12. Write a `Planned` fingerprint row for every briefed post Key into `memory/phase4-叙事指纹索引.md` (`DraftHash: PENDING`), including the two `PrewritingControlFields` (`AssignedFlair`, `PostingTrigger`).
+13. Write or update the memory files listed below, then show the Brief gate checklist.
 
 ## Outputs
 
@@ -97,9 +109,10 @@ Optional Feishu Brief documents require an explicit user-provided folder link.
 
 Writing may start only after:
 
-- Every target subreddit has a valid, unexpired baseline card with a `SampleSetID`.
+- Every target subreddit has a valid, unexpired baseline card with a `SampleSetID`, with flair coverage recorded (`BaselineScope` / `FlairCoverageStatus`); `FlairCoverageStatus` is `Exact` or `ApprovedFallback` (never `Insufficient`).
+- Every **main-post** Key carries a non-empty `AssignedFlair` — drawn from the subreddit's live flair list, or the sentinel `NoFlairAvailable` / `NoFlairRequired` for flairless boards — and its `AssignedFlair` matches the baseline card's covered flair (or an approved fallback).
 - Every **main-post** Key's `客户内容原型` resolves to an `Approved` mapping row (`MappingRuleID` / `MappingVersion` recorded in the Routing Decision). Self-comments inherit the paired post's value; external comments use `—` and need no mapping of their own.
-- Every post Brief row carries the nine `LockedFingerprintFields` (`ContentFunction`, `StructureFamily`, `OpeningMove`, `EvidenceShape`, `BrandEntryMode`, `CaveatMode`, `EndingMode`, `LengthBand`, `AnchorSlot`) and the four `TraceFields` (`SampleSetID`, `RecentCollisionCheck`, `MappingRuleID`, `MappingVersion`).
+- Every post Brief row carries the nine `LockedFingerprintFields` (`ContentFunction`, `StructureFamily`, `OpeningMove`, `EvidenceShape`, `BrandEntryMode`, `CaveatMode`, `EndingMode`, `LengthBand`, `AnchorSlot`), the two `PrewritingControlFields` (`AssignedFlair`, `PostingTrigger` — the latter satisfying the `PostingTrigger × BrandEntryMode` cross-constraint), and the four `TraceFields` (`SampleSetID`, `RecentCollisionCheck`, `MappingRuleID`, `MappingVersion`).
 - `RecentCollisionCheck` is `Pass` for every row, or `Exception` with an approved `ExceptionRef`.
 - `memory/phase4-叙事指纹索引.md` has exactly one active `Planned` row for every briefed **post** Key (comments are out of scope while `WindowScope: PostOnly`).
 - `memory/phase4-撰写brief.md` exists and covers every planned Key for the batch.
@@ -121,8 +134,12 @@ Writing may start only after:
 - BrandSkuPlacementChecked: Yes / No
 - BaselineSampleSetIDs:
 - BaselineStatus: Fresh / Incomplete / Expired / Missing
+- AssignedFlairCoverage: <n>/<N>            # main-post Keys with a non-empty AssignedFlair
+- FlairCoverageStatus: Exact / ApprovedFallback / Insufficient   # batch-level = worst status across cards (Insufficient > ApprovedFallback > Exact)
 - ClientPrototypeMappingStatus: AllApproved / BlockingGap
 - StructureFieldsAssigned: Yes / No
+- PrewritingControlFieldsAssigned: Yes / No   # AssignedFlair + PostingTrigger per main-post Key
+- PostingTriggerCrossConstraintOK: Yes / No    # UnsolicitedTestimony ⇒ BrandEntryMode 不出现
 - FingerprintPlannedRows: <n>/<N>
 - DiversityWindow: WindowCount=<n>, WindowDays=<n>
 - RecentCollisionCheck: Pass / BlockingCollision / Exception(<ExceptionRef>)

package/skills/reddit-phase-brief/references/phase4-brief-workflow.md

@@ -133,9 +133,13 @@ Brief 去重检查报告：
 > 位置、结尾形态。放在写作阶段才采集，等于所有风格规则先用跨板块默认值填一遍，
 > baseline 事后只能微调语气——这正是 v2.12.24 之前 baseline 形同虚设的根因。
 
-**对本批每一个目标 subreddit** 建立或复用一张 baseline 卡：
+**W0 前置（先于取样）：** 为本批每个主帖 Key 确定 `AssignedFlair`（取自该 subreddit
+现行 flair 列表，不得自造；缺失 → `BlockingGap`）。baseline 采样口径**优先按
+`AssignedFlair` 细分**，因为 flair 往往同时决定发帖资格、披露与帖型，不是纯风格标签。
 
-- 同 subreddit、同 flair / 同内容功能的 **5–10 条**近期真实帖（附 URL）
+**对本批每一个目标 subreddit（并优先按 `AssignedFlair` 细分）** 建立或复用一张 baseline 卡：
+
+- 同 subreddit、**同 `AssignedFlair`**（退而求其次同内容功能）的 **5–10 条**近期真实帖（附 URL）
 - word count **中位数 + IQR**（不是单一范围）
 - 段落数中位数
 - 标题长度与标题句式分布
@@ -152,8 +156,23 @@ Brief 去重检查报告：
 - `BaselineMaxAgeDays:`（默认 60，取自指纹索引文件头）
 - `ValidUntil:` = `AcquiredAt + BaselineMaxAgeDays`
 - `BaselineStatus:` `Fresh` / `Incomplete`（样本够但维度不全）/ `Expired` / `Missing`
-  / `TruncatedSample`（样本正文被截断，长度类维度不可信）
-  —— 只有 `Fresh` 允许进入写作；`Incomplete` 必须补维度，**不得用默认值顶替**
+  —— 与 shared 契约同为四值闭集；只有 `Fresh` 允许进入写作；`Incomplete` 必须补维度，
+  **不得用默认值顶替**。样本正文被截断（长度类维度不可信）按 `Incomplete` 处理并注明
+  `IncompleteReason: TruncatedSample`，**不新设第五个 `BaselineStatus` 值**
+
+flair 覆盖字段（与 `BaselineStatus` 正交，权威定义见
+`../reddit-strategy-shared/references/content-diversity-contract.md`）：
+
+- `BaselineScope:` `ExactFlair` / `FlairCluster` / `SubredditFallback`
+- `BaselineFlair:` 该卡实际覆盖的 flair（`ExactFlair` 时等于 `AssignedFlair`）
+- `FlairSampleCount:` 精确 `AssignedFlair` 的合格样本数
+- `FlairCoverageStatus:` `Exact` / `ApprovedFallback` / `Insufficient`
+- `FallbackApprovalRef:` 用 cluster / subreddit fallback 时的批准署名
+
+优先取 exact-flair 5–10 条；不足时先合理扩大时间窗重采（不得混入其他 flair 仍称
+exact）。仍不足时二分：flair 决定资格 / 披露 / 帖型 / 上下文 → `Insufficient` +
+`BlockingGap`，不得 fallback，回流 `reddit-subreddit-compliance`；仅缺风格分布且有
+相邻 flair / 板块主样本 → 显式 `ApprovedFallback` 并记 `FallbackApprovalRef`。
 
 **样本资格门禁（v2.12.30 新增，硬规则）：** baseline 卡的 5–10 条样本必须是
 **真实用户帖**。`reddit-voc-volume` 逐条输出 `sample_role` 与 `body_status`，
@@ -380,6 +399,16 @@ ASK / GUIDE / ROUNDUP / OWNER / OPINION / RANT / FOLLOWUP / LIST / REPUTATION
    `EndingMode` / `LengthBand`（由该 subreddit baseline 的中位数 + IQR 派生）/
    `AnchorSlot`（`ABURequired=Yes` 时须在该 family 的 `AllowedAnchorSlots` 内；
    否则填 `N/A`）。
+   **并预分配两个前置控制字段**（`PrewritingControlFields`，权威定义见
+   `../reddit-strategy-shared/references/content-diversity-contract.md`）：
+   - `AssignedFlair`：已在 B-1.0/W0 确定（取自该 subreddit 现行 flair 列表；板块无
+     flair 或允许不选时填哨兵值 `NoFlairAvailable` / `NoFlairRequired`，不得留空、
+     不得用评论的 `N/A`），此处只核对与 baseline 覆盖一致。
+   - `PostingTrigger`：受控值 `ConcreteFailure` / `PurchaseDecision` /
+     `MeasurementResult` / `CompatibilityGap` / `UnsolicitedTestimony`。满足
+     `PostingTrigger × BrandEntryMode` 交叉约束：`BrandEntryMode ≠ 不出现` 禁取
+     `UnsolicitedTestimony`；取 `UnsolicitedTestimony` 时 `BrandEntryMode` 必须
+     `不出现` 且有真实非征集非品牌控制的 testimony 证据。
 6. **`RecentCollisionCheck`**（必须在第 5 步之后，因为需要完整三元组）：读取该
    subreddit 的有效窗口（`WindowCount` 且 `WindowDays` 同时满足，按 `WindowSort`
    排序，**排除当前 Key 自己那一行**），检查本篇
@@ -413,6 +442,10 @@ ASK / GUIDE / ROUNDUP / OWNER / OPINION / RANT / FOLLOWUP / LIST / REPUTATION
 ```markdown
 ### [Key] · [主帖型代码] · [P-B✅ / P-NB❌] · 绑定 SKU：[SKU]
 
+**PrewritingControlFields（前置控制字段，主帖必填，先于 baseline 与结构）：**
+- AssignedFlair:          （现行 flair 之一；无 flair 板块填 NoFlairAvailable / NoFlairRequired，不留空）
+- PostingTrigger:         （ConcreteFailure / PurchaseDecision / MeasurementResult / CompatibilityGap / UnsolicitedTestimony；须满足 × BrandEntryMode 交叉约束）
+
 **LockedFingerprintFields（v2.12.26 必填 9 项锁字段，写大纲之前先填满）：**
 - ContentFunction:
 - StructureFamily:
@@ -544,6 +577,8 @@ ASK / GUIDE / ROUNDUP / OWNER / OPINION / RANT / FOLLOWUP / LIST / REPUTATION
 [ ] 【v2.12.26】项目差异化窗口已锁定（WindowCount 8–12 且 WindowDays 30–60 同时生效 / WindowScope: PostOnly / WindowSort / BaselineMaxAgeDays / IndexRevision），写入指纹索引文件头
 [ ] 【v2.12.26】每个**主帖** Key 的「客户内容原型」均命中 memory/phase0-客户内容原型映射.md 的 Approved 行，MappingRuleID 与 MappingVersion 已记入 Routing Decision；无未映射代号（自评继承配对主帖的原型值，外部评论填「—」，不单独要求映射）
 [ ] 【v2.12.26】每条主帖 Brief 均含 9 项 LockedFingerprintFields：ContentFunction / StructureFamily / OpeningMove / EvidenceShape / BrandEntryMode / CaveatMode / EndingMode / LengthBand / AnchorSlot
+[ ] 【P0-C】每条主帖 Brief 均含 2 项 PrewritingControlFields：AssignedFlair（非空，现行 flair 或 NoFlairAvailable/NoFlairRequired）+ PostingTrigger（受控五值，满足 × BrandEntryMode 交叉约束）
+[ ] 【P0-C】每张 baseline 卡记录 flair 覆盖：BaselineScope / BaselineFlair / FlairSampleCount / FlairCoverageStatus（Exact 或 ApprovedFallback；Insufficient → BlockingGap）/ FallbackApprovalRef（fallback 时须署名）
 [ ] 【v2.12.26】每条主帖 Brief 均含 4 项 TraceFields：SampleSetID / RecentCollisionCheck / MappingRuleID / MappingVersion
 [ ] 【v2.12.26】每条主帖 Brief 均含 Routing Decision 块，五层候选收敛轨迹齐全（Baseline / Function / Evidence / PrimaryAndCompliance / MappingPredicates + MatchedPredicate），SelectedCandidateRank 与 SkippedCandidates（CandidateRank + RejectionReason）已记录；跳过更靠前候选的理由均为 1–5 层硬失败，差异化仅作 tie-breaker，CollisionRepairPath 已记录
 [ ] 【v2.12.26】RecentCollisionCheck 全部为 Pass，或为 Exception 且有已批准的 ExceptionRef（例外须引用 SampleSetID + 样本占比 + 批准人）

package/skills/reddit-strategy-shared/references/content-diversity-contract.md

@@ -53,6 +53,8 @@ LastUpdatedAt: <ISO timestamp>
 | `DraftRef` | 指向终稿位置（如 `memory/phase4-writing-batch-2.md#PLAUD-P007`） |
 | `ExceptionRef` | 有例外时指向例外记录；不得在单元格里写一句理由了事 |
 | `HashInput` | 本索引 `WindowScope: PostOnly`，因此这里恒为 `CanonicalPostPayload`。`CanonicalCommentPayload` 只出现在 `memory/content-review-log.md`，不出现在本索引 |
+| `AssignedFlair` | 该主帖 Key 的目标发布 flair（`PrewritingControlField`）。Brief 预分配、writing 前校验、delivery 反查其与 baseline 覆盖关系。为空即 `BlockingGap`（评论 / 回复不适用） |
+| `PostingTrigger` | 「这篇帖为什么现在值得发」的现实触发（`PrewritingControlField`，受控值见下）。Brief 预分配、writing 从终稿反推并与 `Planned` 值核对；漂移即阻断。**不并入硬碰撞三元组** |
 
 ### 记录状态机
 
@@ -101,6 +103,14 @@ LastUpdatedAt: <ISO timestamp>
 | `LengthBand` | 依据该 subreddit baseline 的中位数 + IQR 派生的档位，**不设跨板块固定词数** |
 | `AnchorSlot` | 见 ABU 七种合法位置；无需 ABU 时填 `N/A` |
 
+**两个前置控制字段（`PrewritingControlFields`，与 9 个 `LockedFingerprintField`
+并列但不并入锁字段集，也不进硬碰撞三元组）：**
+
+| 字段 | 受控值 |
+|---|---|
+| `AssignedFlair` | 该主帖 Key 的目标发布 flair（**受控于该 subreddit 的现行 flair 列表**，不得自造）。主帖必填、不得留空；板块无 flair 或允许不选时用哨兵值 `NoFlairAvailable` / `NoFlairRequired`（不得与评论的 `N/A` 混用）。评论 / 回复填 `N/A` |
+| `PostingTrigger` | `ConcreteFailure` / `PurchaseDecision` / `MeasurementResult` / `CompatibilityGap` / `UnsolicitedTestimony`（闭集）。回答「这篇帖为什么现在值得发」 |
+
 字段语义补充：
 
 - `EvidenceShape` 必须记录**终稿实际使用**的证据形态，不得记录「计划用但正文里没有」。
@@ -125,6 +135,24 @@ LastUpdatedAt: <ISO timestamp>
   - `CaveatMode` 回答「限制**以什么形式**出现在正文里」——它是表达形态。
   - `RealLimitationStatus: NoneKnown` 时 `CaveatMode` 必须是 `无真实限制则不写`；
     `EvidenceRequired` / `BlockingUnknown` 时不得先写死 `CaveatMode` 再去补证据。
+- `AssignedFlair` / `PostingTrigger` 是 **Brief 阶段预分配的前置控制字段**，不是
+  写手临场自选，也不是终稿事后补记：
+  - `AssignedFlair` 决定 baseline 采样口径、subreddit-fit 判定基准，往往同时决定
+    发帖资格与披露要求。改动 `AssignedFlair` 使原 flair baseline 失效，须退回
+    `reddit-phase-brief` 重新确认 baseline + 合规，再重跑结构适配。它**不是**第 10 个
+    `LockedFingerprintField`（不是正文结构指纹），但享有同级别的变更纪律。
+  - `PostingTrigger` 与 `ContentFunction`（帖子完成什么任务）、`EvidenceShape`
+    （正文用什么证据形态）互不等价：它描述**触发本次表达的现实事件或决策节点**。
+    Brief 预分配、writing 从终稿反推并与 `Planned` 值核对，`Planned` 与实际不符即
+    `BlockingGap`（等同锁字段漂移处理，`FixOwnerSkill: reddit-phase-brief`）。
+  - **`PostingTrigger` × `BrandEntryMode` 交叉硬约束（双向）：**
+    - `BrandEntryMode ≠ 不出现` 时，`PostingTrigger` **禁取** `UnsolicitedTestimony`。
+    - 反之，`PostingTrigger = UnsolicitedTestimony` 时，`BrandEntryMode` **必须**是
+      `不出现`；且 `SourceEvidence` 必须存在真实、非征集、非品牌控制的 testimony，
+      正文不得冒充该 testimony 的原作者。两者任一不满足即 `BlockingGap`。
+  - 两个控制字段都**不参与**硬碰撞三元组（`StructureFamily + OpeningMove +
+    EndingMode`）；把它们并进去会稀释碰撞门禁（换个 trigger 标签即可绕过结构重复）。
+    它们只做 Planned-vs-final 漂移校验与上面的交叉约束。
 - 本索引与硬碰撞规则的作用域是 `PostOnly`：**只有主帖 Key 需要指纹行**。
   评论 / 回复不进入本索引，也不需要 `Final` 行；它们仍然需要 Hub 审核与
   `FinalDraftSha256`。评论若需防同质化，另行按
@@ -251,8 +279,31 @@ FixOwnerSkill: reddit-phase-brief
 
 只有 `Fresh` 允许进入写作。
 
-- 同一 subreddit、同一有效期内的多个 Key **复用同一个 `SampleSetID`**，baseline
-  不必每帖重抓。
+**flair 覆盖是与 `BaselineStatus` 正交的第二维（不要塞进上面的四值枚举）。**
+baseline 采样口径优先按 `AssignedFlair` 细分：
+
+| 字段 | 取值 / 含义 |
+|---|---|
+| `BaselineScope` | `ExactFlair`（精确 flair 5–10 条合格样本）/ `FlairCluster`（相邻 flair 合并，须批准）/ `SubredditFallback`（无 flair 细分，用板块主样本，须批准） |
+| `BaselineFlair` | 该 baseline 实际覆盖的 flair；`ExactFlair` 时等于 `AssignedFlair` |
+| `FlairSampleCount` | 精确 `AssignedFlair` 的合格样本数 |
+| `FlairCoverageStatus` | `Exact` / `ApprovedFallback` / `Insufficient` |
+| `FallbackApprovalRef` | 用 cluster / subreddit fallback 时的批准记录（策略师 / 用户署名） |
+
+flair 覆盖门禁（不是「每个 flair 不足 5 条一律阻断」）：
+
+1. 先取 exact-flair 5–10 条 → `BaselineScope: ExactFlair` / `FlairCoverageStatus: Exact`。
+2. 精确样本不足时先合理扩大时间窗口重采；**不得混入其他 flair 后仍声称 exact**。
+3. 仍不足时按缺口性质二分：
+   - flair 决定**发帖资格 / 披露 / 帖型 / 必备上下文** → `FlairCoverageStatus: Insufficient`，
+     `BlockingGap`，**不得 fallback**，回流 `reddit-subreddit-compliance`。
+   - 只缺**风格分布**、且有相邻 flair 或板块主样本 → 允许显式 `FlairCluster` /
+     `SubredditFallback`，记 `FallbackApprovalRef` 与不确定性，`FlairCoverageStatus: ApprovedFallback`。
+4. `BaselineStatus` 继续只表达完整性 / 新鲜度（`Fresh`/`Incomplete`/`Expired`/`Missing`），
+   flair 覆盖状态一律走 `FlairCoverageStatus`，两者不得混算。
+
+- 同一 subreddit、同一有效期、**同一 flair 覆盖口径**内的多个 Key **复用同一个 `SampleSetID`**，
+  baseline 不必每帖重抓；`AssignedFlair` 不同且需 exact 覆盖时不得跨 flair 复用。
 - `SampleSetID` 缺失或已过期 → `BlockingGap`（详见 `reddit-phase-brief`）。
 - 单条任务用到的产品事实、VOC 案例、Geoly 引用放在 Brief 的
   `RoutingEvidenceIDs`，**不要塞进 `SampleSetID`**。
@@ -393,8 +444,8 @@ FingerprintSyncStatus: PendingWritingOwner
 
 | 层 | 执行者 | 内容 | 失败行为 |
 |---|---|---|---|
-| 1 · 预分配 | `reddit-phase-brief` | 在 Brief 里预先锁定 9 个结构字段并写入 `Planned` 指纹行；**不让 writer 临场自选** | Brief 缺字段 → `BlockingGap`，不得进入写作 |
-| 2 · 写作硬门禁 | `reddit-content-writing` | 开写前必须读取有效窗口并完成碰撞检查；写完后从终稿推导实际指纹 | 未读窗口 / 未做碰撞检查 / 实际与 `Planned` 结构字段不符 → 停止并退回 Brief |
+| 1 · 预分配 | `reddit-phase-brief` | 在 Brief 里预先锁定 9 个结构字段 + 2 个 `PrewritingControlField`（`AssignedFlair` / `PostingTrigger`）并写入 `Planned` 指纹行；**不让 writer 临场自选** | Brief 缺字段（含 `AssignedFlair` / `PostingTrigger`）→ `BlockingGap`，不得进入写作 |
+| 2 · 写作硬门禁 | `reddit-content-writing` | 开写前必须读取有效窗口并完成碰撞检查；写完后从终稿推导实际指纹与 `PostingTrigger` | 未读窗口 / 未做碰撞检查 / 实际与 `Planned` 结构字段或 `PostingTrigger` 不符 → 停止并退回 Brief |
 | 3 · 交付反查 | `reddit-delivery-check` | 反向校验窗口无碰撞、映射已批准、结构未被 rewrite 改动、`AnchorSlot` 合法 | 任一不符 → `OverallStatus: NeedsFix` |
 | 4 · hash 绑定 | `reddit-delivery-check` | 重算终稿 hash，与指纹行、review log 三方比对 | 任意两者不一致 → 阻断交付 |
 
@@ -431,6 +482,10 @@ FingerprintSyncStatus: PendingWritingOwner
 - Hub 返回的 `optimized` 改变了任一受保护结构字段而被整稿采纳
   （`HubStructureConflictStatus: Escalated` 未解决）；
 - 终稿已修改却仍沿用旧的 Hub pass；
+- 主帖 Key 的 `AssignedFlair` 为空，或 `FlairCoverageStatus: Insufficient`（flair
+  决定资格 / 披露 / 帖型 / 上下文却无合格样本、又不满足 `ApprovedFallback`）；
+- `PostingTrigger` 的 `Planned` 值与终稿反推值不符，或违反
+  `PostingTrigger × BrandEntryMode` 交叉约束；
 - 客户原型未映射（见 `content-type-routing-contract.md`）。
 
 固定输出：
@@ -438,7 +493,7 @@ FingerprintSyncStatus: PendingWritingOwner
 ```markdown
 FingerprintStatus: BlockingGap
 ContentKey: <Key>
-MissingRequirement: <index | window | planned row | final hash | sample set | mapping>
+MissingRequirement: <index | window | planned row | final hash | sample set | mapping | assigned flair | flair coverage | posting trigger>
 AllowedToWrite: No
 AllowedToDeliver: No
 RequiredAction: <Brief repair | fingerprint sync | final draft re-hash>

package/skills/reddit-strategy-shared/references/evidence-policy.md

@@ -24,6 +24,36 @@ Use this file to resolve conflicts between Reddit strategy sources. If older pla
 - Only `reddit-subreddit-compliance` output decides subreddit eligibility before planning; `reddit-voc-volume` never issues the compliance verdict itself.
 - Phase 4 execution table is the writing truth after it is locked. If Brief or draft writing needs to change execution-table fields, return to Phase 4.0/4.0.6 and rerun verification.
 
+## Claim Provenance — `ClaimEvidenceStatus`（三态，可发布文本准入判据）
+
+**每个进入可发布文本（标题 / 正文 / 评论 / 回复）的实质性 claim** 必须归入以下三态
+之一。此定义是权威源，`reddit-content-writing` 与 `reddit-content-rewrite` 的 A 层
+直接引用，不各自发明。
+
+| 状态 | 定义 | 可发布文本 |
+|---|---|---|
+| `Cited` | 可映射到项目已批准证据（`KnownProductFacts` / SKU 基准表 / VOC / 甲方语料 / 用户提供材料），并有 `ClaimRef` | 允许 |
+| `ExternallyVerified` | **非产品事实类** claim（法规、公开兼容信息、公开行业数据等），经外部权威来源独立核验，并有 `ExternalSourceRef` | 允许 |
+| `Inferred` | 模型综合、推测或因果推断，无直接来源 | **禁止** |
+
+衔接与边界：
+
+- `Cited` 向后兼容旧规「所有 claim 可追溯到 `KnownProductFacts` / `SourceEvidence`」。
+- `ExternallyVerified` 只覆盖**背景事实**；**产品性能、亲身体验、使用时长**不因为
+  搜到第三方页面就自动成为作者可声称的事实（那类仍须 `Cited` 到已批准证据，或删）。
+- `Inferred` 不是「低置信度正文」，而是**可发布文本禁入状态**：只能留在 Brief 的内部
+  推理区，不得进入标题 / 正文 / 评论 / 回复；若无法删除且为完成内容所必需 → `BlockingGap`。
+- 只对**实质性 claim** 分类；明确的主观判断、常识性连接句不必逐句台账。
+- 最小记录形态（按需，不强制逐句）：
+
+  ```markdown
+  ClaimLedger:
+  - Claim:
+    ClaimEvidenceStatus: Cited | ExternallyVerified | Inferred
+    ClaimRef:            # Cited → 证据 ID；ExternallyVerified → ExternalSourceRef
+    PublishableTextAllowed: Yes | No   # 标题 / 正文 / 评论 / 回复皆适用
+  ```
+
 ## Known v2.12.7 Corrections
 
 - Reddit Phase 0 does not run the full `kg-action-report` workflow. Use `kg-reddit-check` for positioning and narrative reading.