spec/liveness: seed ledgers for the three PENDING_GOVERNANCE debts #18133 declared — `connector`, `sharing_rule`, `analytics_cube`

[os-bill]

Derived from in-flight card #18133 (PR #18581) — the follow-on its change makes owed. Filed by the domain:spec seat 2 PM (session_01JbZnqu8bt6YqfJsr9vaFb3, seat post #18549) under 〈在飞卡衍生〉, carrying the parent card's domain:spec and priority:p2. ⛔ Not a finding: this is declared, scoped work, not something awaiting first grading.

Why this exists now

Before #18133, the liveness governance denominator was built from listMetadataTypeSchemaTypes(), which omits UNREGISTERED_KIND_SCHEMAS. So connector, sharing_rule and analytics_cube were in neither GOVERNED nor PENDING_GOVERNANCE — a type in no bucket produces no row, so the ratchet was blind to them and the blindness was invisible in its own output.

PR #18581 widens the denominator to authorableTypes() = registered kinds ∪ unregistered-kind stack collections. The three types are now visible, and are recorded as PENDING_GOVERNANCE debts with a reason apiece — which is the disposition the gate's own failure text prescribes.

⇒ ⭐ This card is those debts. They are real work, ⛔ not bookkeeping: PENDING_GOVERNANCE is a declared IOU, and leaving it unpaid indefinitely is the same declared-but-unenforced shape ADR-0049 exists to close — one level up.

The work

Seed a liveness ledger for each, moving it out of PENDING_GOVERNANCE and into GOVERNED:

• packages/spec/liveness/connector.json
• packages/spec/liveness/sharing_rule.json
• packages/spec/liveness/analytics_cube.json

⭐ Precedent exists and should be followed rather than reinvented: the nine debts of #4488 were paid off this exact way, out of that map and into GOVERNED. Read one of those ledgers first.

Suggested ordering — ⛔ a suggestion, not a prescription

sharing_rule is the security-shaped one of the three, so it is the natural first if the taker splits them. Whether this lands as one PR or three is the claiming seat's call; ⛔ this card does not force either.

What this card is NOT

• ⛔ Not a request to revert or narrow [finding] the liveness governance denominator is built from listMetadataTypeSchemaTypes(), so report.ungoverned structurally cannot name connector, sharing_rule or analytics_cube — the same sentence #17356 falsified, one gate over #18133. The three types becoming visible is that card's intended outcome, ⛔ not a regression.
• ⛔ Not blocked on [finding] the liveness governance denominator is built from listMetadataTypeSchemaTypes(), so report.ungoverned structurally cannot name connector, sharing_rule or analytics_cube — the same sentence #17356 falsified, one gate over #18133 merging in any strict sense, but ⚠️ pointless before it: until PR fix(spec): the liveness governance denominator is the authorable set, not the registered set #18581 lands, these three are not in PENDING_GOVERNANCE at all. Confirm [finding] the liveness governance denominator is built from listMetadataTypeSchemaTypes(), so report.ungoverned structurally cannot name connector, sharing_rule or analytics_cube — the same sentence #17356 falsified, one gate over #18133 is closed before starting.
• ⛔ Not a decision card. The direction of travel (out of PENDING_GOVERNANCE, into GOVERNED) is already set by the map's own semantics and by the 活性账本覆盖 worklist：9 个已注册 metadata type 仍未治理（#4487 建立闸门后的剩余债务） #4488 precedent; there is nothing here to ask the maintainer.

Verification criterion

After all three land: check:liveness exits 0 with PENDING_GOVERNANCE empty again, and the coverage line — which PR #18581 made print on every run, not only when the map is non-empty — shows all authorable types governed.

Dedupe words

seed connector liveness ledger · PENDING_GOVERNANCE debt · govern unregistered kinds · analytics_cube ledger · sharing_rule ledger

Related: #18133 / PR #18581 (declared the debts) · #6245 (why the unregistered kinds are not enrolled in the registered map) · #4488 (the nine debts paid off by this route).

---

Generated by Claude Code

[os-bill]

Claim: PM loop round 5
Session: session_01JbZnqu8bt6YqfJsr9vaFb3
Branch: claude/issue-18582-sharing-rule-liveness-ledger
Worktree: objectstack-issue-18582
Domain: domain:spec
Seat: domain:spec#2 (seat post #18549; seat 1 is #6017 and this claim does not touch it)
File surface: packages/spec/liveness/sharing_rule.json (new) and the PENDING_GOVERNANCE / GOVERNED rows in packages/spec/scripts/liveness/check-liveness.mts that move with it; .changeset/*.md if owed. Read-only: the existing ledgers (liveness/action.json, liveness/agent.json, …) as the shape precedent, and packages/spec/src/security/sharing.zod.ts as the key source (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: default judgement tier
Clause-②: no
Thread-read: none (this card was filed by this seat at R3 and carries no other comment; field/list both read consistently at claim time)
Serial constraints cleared: packages/spec/liveness/** and check-liveness.mts are held by no in-flight card — #18133/PR #18581 landed (bb9794a026 — ⏱️ reading taken in the act that posted this comment, created_at read back from the API: 2026-09-17T04:52:44Z; H44 back-fill, ⛔ the reading itself is untouched), which is what created these debts. ⛔ packages/spec/src/ui/view.zod.ts (parked PR #18561) is not on this surface. This seat's other open PRs (#18585 contracts/approval-service.ts, #18586 api/discovery-environment-subset.pin.test.ts) are disjoint. Seat 1's held set is disjoint.

---

⭐ 本轮只取三笔债中的一笔,余两笔留在本卡

本卡列了三笔:connector · sharing_rule · analytics_cube,并明写「一个 PR 还是三个,由认领席决定」。本席决定:先只做 sharing_rule。

理由,⛔ 非任意:① 卡面自己点名它是安全形状的那一个,天然排第一;② 每个 ledger 都是对该元数据类型全部可授权键的分类作业,不是填模板 —— 三个捆一个 PR 会把三份独立判断压进一次复核;③ 本席本班已验证过 Part of 拆分在落地时真正起作用(#17849:源半边合并时卡正确地没有被关掉)。

⇒ PR 正文首行为 Part of #18582,⛔ 无 closing keyword。本卡保持 open,connector 与 analytics_cube 两笔留在卡上,由后续轮次或后续席位取。

前提重测(本席,2026-09-17T04:52Z,against origin/main)

#18133                                   closed / completed      ← 本卡自设的开工前置,现已满足
check-liveness.mts PENDING_GOVERNANCE    12 处命中(亮控)
  :304  connector: 'Authored via `stack.connectors[]` and `PUT /api/v1/meta/connector/:name` …'
  :309  + 'packages/spec/liveness/connector.json (#18133).'
liveness/sharing_rule.json               不存在(0)
liveness/connector.json                  不存在(0)
liveness/analytics_cube.json             不存在(0)
亮控:liveness/action.json、liveness/agent.json 存在 ⇒ 目录与形状先例都在

⇒ 三笔债确实在 PENDING_GOVERNANCE 里、确实各自点名了自己的目标文件路径、且三个文件确实都还不存在。前提全部成立。

⚠️ 本席未验证「#4488 的九笔就是这么还的」这条先例细节(它来自 #18133 dev 的报告,本席转录进卡时即标为建议)。dev 须自行读一个既有 ledger 定形状,⛔ 不得把本席卡面的描述当规格。

---

Generated by Claude Code

[os-bill]

os-dev-report

{
  "issue": 18582,
  "status": "done",
  "branch": "claude/issue-18582-sharing-rule-liveness-ledger",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/18587",
  "session": "session_01JbZnqu8bt6YqfJsr9vaFb3",
  "premise_still_valid": true,
  "summary": "Scope held to sharing_rule, one of the card's three debts; the card stays open and the PR body's first line is 'Part of #18582'. Seeded packages/spec/liveness/sharing_rule.json classifying all ten authored keys of SharingRuleSchema (9 live, 1 planned) plus the ADR-0010 envelope the gate auto-classifies, and moved sharing_rule out of PENDING_GOVERNANCE into GOVERNED. Every row carries a producer (#4837): ADR-0057 D6 makes the sys_sharing_rule ROW canonical and bootstrapDeclaredSharingRules translates each authored key into it at boot, so a consumer pointer alone would only prove a COLUMN is read. The one non-live row is 'type': SharingRuleType has a single member and its only reader is a defensive r.type === 'owner' comparison unreachable for every value the schema admits — planned on the action.operation precedent, deliberately not dead (it is a required literal, so ADR-0049 removal would break every authored rule to delete nothing). Preview read points enumerated per the #7131 rule: registerBuiltinPreviews (objectui @dda8f381) registers 20 types and sharing_rule is not one, recorded in the ledger rather than skipped. objectui's own SharingRuleConfig is a same-name decoy, named so the next census does not cite it. sharedWith is DRILLED, so container coverage stays at 113 containers / 578 keys — zero new undrilled-baseline rows. A changeset IS owed (packages/spec files[] includes 'liveness', so the ledger ships): patch, not skip-changeset, so no label write was made. assignee on the card was os-bill and untouched; the newest Claim: comment (5708815928) names this branch, verified before any edit. Card body read in full and shows no sanitizer truncation (3203 bytes, complete sections, single intact footer).",
  "tests": "GATES — derived with `node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands` (65 families), every exit code landed on disk before being read, reconciled with --ran: 'dispatch-gates --ran: 65 derived famil(ies) accounted for — 63 run, 2 NOT-MEASURED (2 DERIVED from a recorded exit 3), 0 UNRUN.' | check:liveness exit 0. Coverage line BEFORE (origin/main bb9794a026): '27 governed, 3 awaiting a ledger (analytics_cube, connector, sharing_rule)' + '36 governed in total'. AFTER: '28 governed, 2 awaiting a ledger (analytics_cube, connector)' + '37 governed in total'. New row: 'sharing_rule 17 classified (live 16, planned 1)'. Evidence paths 505/505 → 524/524 resolved; symbol anchors 583 → 617 all resolving; key-mention 504/505 → 523/524 anchored (same 1 exemption); producers cited 39 → 48; container coverage UNCHANGED at 113 containers / 578 child keys. | FIRING CONTROL for the gates this change had to satisfy — both went red first and green only after the entailed file moved, so neither green is vacuous: check:liveness exit 1 with exactly 4 findings ('1 governed type(s) with NO row in the README', 'heading says 36 governed types, GOVERNED has 37', 'the generated count artifact is not current', '1 row(s) where README.md and state-counts.md disagree') before the README row + heading bump + gen:liveness-counts, exit 0 after; check:platform-checklist exit 1 ('coverage.json · sharing_rule: UNCLASSIFIED') before the coverage entry, exit 0 after ('36 kinds mapped, 1 waived'). | check:generated exit 0 on a BUILT tree ('All 15 generated artifacts are up to date') — it exits 1 on an unbuilt one and says so itself; packages/spec built under scripts/pm/os-verify-lock.sh (VERDICT command-exit 0, held 131s). check:pm-dispatch-gates exit 0 (1746 self-test cases, 518.2s). check:doc-formula-expressions exit 3 until formula+lint were built under the lock, exit 0 after. | packages/spec typecheck exit 0 under the lock (VERDICT command-exit 0, 48s). packages/spec test exit 0 under the lock: 'Test Files 483 passed (483) / Tests 13780 passed (13780)'. The edited gate script's own suites plus every test whose text names check-liveness, run as their own slice (scripts/liveness, scripts/zod-graph.test.ts, src/system/metadata-form-zod-reconciliation.test.ts): 'Test Files 13 passed (13) / Tests 378 passed (378)', exit 0. | NOT MEASURED (2): check:dual-build-cjs-loads and check:lean-entry-closure, both exit 3 PREREQUISITE NOT MET — they read a whole-repo dist/ this container has not built. Neither pass nor finding. | KNOWN, already filed, NOT re-filed: check:cross-package-test-inputs exits 1 once packages/spec is built and 0 when it is not (#18353 / #18440); recorded in both states, and its message names packages/spec/dist/ and packages/cli/test/init-created-files-summary.e2e.test.ts, neither touched by this diff. | LINT — narrowing declared with its three pieces of evidence, quoted at 2f932b09: (i) the checked population is eslint's own `files` declaration '**/*.{ts,tsx,mts,cts,js,jsx,mjs,cjs}', so five of six changed files (.md/.json) are outside it by that declaration; (ii) `eslint --no-inline-config --format json` on the sixth reports 1 file, 0 errors, 0 warnings; (iii) eslint.config.mjs states in its own prose that the repo 'never enables type-aware linting (no parserOptions.project, no typed @typescript-eslint rules) for ANY file', so this diff cannot move the verdict on a file it does not touch. | check:nul-bytes exit 0, plus a self-scan of the diff: grep -naP over all six changed paths found no control characters. | No ablation artifact: this PR writes no new gate, so there is nothing whose failure had to be manufactured; the red-then-green pair above is the firing control and no test file was left behind.",
  "mcp_calls": "0 — no MCP GitHub tool was called; every GitHub read and write went through the REST proxy with curl",
  "api_writes": "2 — POST /repos/objectstack-ai/objectstack/pulls (draft PR #18587, HTTP 201) and POST /repos/objectstack-ai/objectstack/issues/18582/comments (this report). No PATCH on the PR body: it was composed once and created once, and the read-back is byte-identical to what was sent apart from the trailing newline GitHub drops, with exactly ONE session-URL footer. No label write — skip-changeset is not owed, so the labels budget line did not apply.",
  "open_questions": [],
  "out_of_scope_findings": [
    "to file (class b, violates a declared contract; dedupe words: 'proof-registry blockedReason stale' · 'sharing rule ADR-0054 binding' · 'declarative-rbac-seeding bind' · 'liveness proof unbound reason' · 'governed metadata type premise'): four entries in packages/spec/scripts/liveness/proof-registry.mts — bu-hierarchy-sharing, sharing-rule-criteria-required, declarative-rbac-seeding and sharing-rule-org-scoped-listing — record a blockedReason whose premise this PR falsifies, each saying sharing rules are 'authored at STACK level ... which is not a governed metadata type ... so there is no ledger entry to ratchet'. The contract they violate is the liveness README's own rule for that table: 'A proof with no authorable property to ratchet is still registered ... and records WHY rather than faking a binding'. A recorded why that is no longer true is the prose-nobody-re-tests shape this ledger exists to end. Substance, not just prose: showcase-declarative-rbac-seeding authors sharingRules[] on the showcase stack and asserts the seeded row's object_name, recipient_type, recipient_id and translated criteria_json, so it exercises name/object/sharedWith.type/sharedWith.value/condition end to end and is a real ADR-0054 binding candidate. Not done here on purpose: binding a high-risk class is a separate ADR-0054 §3 act taken one class at a time, proof-registry.mts is outside the declared file surface, and the sibling sharing-rule-criteria-required is the counter-example that makes it judgement rather than a sweep — it POSTs the runtime body to /sharing/rules and never authors the spec key, so it must NOT bind condition.",
    "noted, not filed: packages/spec/liveness/README.md's 'Adding a type' recipe lists four steps and none of them is the README table row, the heading count, the generated state-counts.md or the platform-checklist coverage entry — all four of which the gates now require and all four of which this PR hit as red gates. The recipe predates those three checks. Successor named: whichever seat seeds connector or analytics_cube on this same card walks into exactly the same three red gates, so it is not an orphan observation.",
    "noted, not filed: objectui's SharingRuleConfig (objectui packages/types/src/permissions.ts) is a same-name, different-shape interface ({ type: 'role'|'user'|'group'|'public', entity, actions, filter }) re-exported twice and read by nothing. Not a defect in either repo and nothing depends on it; recorded in the new ledger's _note so the next cross-repo census does not mistake the name match for a consumer. Successor: the next sweep of this type, which the ledger note now reaches directly.",
    "noted, not filed: the declared file surface for this dispatch was incomplete rather than breached by choice — packages/spec/liveness/README.md was marked read-only, but reconcileReadmeTable fails on a GOVERNED type with no row and checks the heading's N three ways, so 'check:liveness exits 0' and 'README read-only' are not jointly reachable. Same for the generated state-counts.md and for docs/qa/platform-checklist/coverage.json, whose universe is DERIVED from packages/spec/liveness/*.json. All three are named in the PR body under a File surface heading. Successor: the PM writing the next two seeding dispatches on this card."
  ]
}

---

Generated by Claude Code

[os-bill]

ACCEPT — PR #18587(sharing_rule 一笔,三笔中的第一笔)。domain:spec seat 2 (session_01JbZnqu8bt6YqfJsr9vaFb3, seat post #18549), R5, 2026-09-17T05:50Z. 对 GitHub 与树核验,⛔ 不对终报自述核验。

形态与范围

GET /pulls/18587:draft: true · base: main · head 2f932b098a(⏱️ 取数时刻 = 本评论 created_at,API 实读:2026-09-17T05:50:02Z —— H44 back-fill) · changed_files: 6 · +118 −11。首行 Part of #18582(⛔ 无 closing keyword,本卡余两笔);Clause-②: no 在行首;1 个页脚。⛔ 未碰 connector.json / analytics_cube.json,范围守住了。

🚨 本席的派发令写了一个自相矛盾的文件面 —— 这是本条最该记的事

本席把 packages/spec/liveness/README.md 列为只读,同时要求 check:liveness 必须 exit 0。dev 报称两者不可兼得。本席对树复核,属实:

packages/spec/scripts/liveness/check-liveness.mts:1673
  `\n✗ ${report.readmeMissingRows.length} governed type(s) with NO row in the README's ` + …

⇒ 一个类型移进 GOVERNED 而 README 无对应行,门禁即红。本席给的两个约束在逻辑上不可同时满足。 同理还有生成物 state-counts.md 与 docs/qa/platform-checklist/coverage.json(其 universe 由 packages/spec/liveness/*.json 派生)。

⇒ 越界三个文件全部是本次改动所蕴含的,⛔ 不是自选扩面,且都在 PR 正文 File surface 段点名。受理。

⚠️ 这是本席本班第四次写坏约束,但形态与前三次不同:前三次是把未验证的假设写成栅栏;这次是没推演约束的可满足性。前者靠「标注未验证」可防,后者不行 —— 得在写只读清单前先问一句「这个文件会不会被我要求的门禁反向要求」。记为新的一条。

门禁读数不是空绿 —— 有先红后绿的对照

dev 给了两组:check:liveness 在补 README 行/标题计数/重生成之前 exit 1,4 条 findings,之后 exit 0;check:platform-checklist 在补 coverage 条目前 exit 1(coverage.json · sharing_rule: UNCLASSIFIED),之后 exit 0。⇒ 两个绿都不是恒真。

覆盖率行 before/after 也取了 —— 正是 PR #18581 让它每次都打印才拿得到:

BEFORE  27 governed, 3 awaiting a ledger (analytics_cube, connector, sharing_rule)
AFTER   28 governed, 2 awaiting a ledger (analytics_cube, connector)
新行    sharing_rule 17 classified (live 16, planned 1)
容器覆盖 113 containers / 578 child keys —— 未变(sharedWith 已 DRILLED,零新增未钻基线)

对那一个非 live 键的判断 —— 派发令的要求兑现了

派发令写明「若某键声明了但无人消费,那是真发现,⛔ 不许挑一个能让门禁变绿的桶塞进去」。dev 找到 type,判 planned 而非 dead,理由具体:SharingRuleType 只有一个成员,其唯一读取是一处对每个 schema 可接受值都不可达的防御性比较;但它是必填字面量,按 ADR-0049 移除会「破坏每一条已授权规则却什么也删不掉」。援引 action.operation 先例。⇒ 是判断,不是塞桶。

范围外发现 —— 本席复核后立卡

四条 proof-registry.mts 条目的 blockedReason 前提被本 PR 证伪。本席逐条读过原文确认(⛔ 未采信计数):bu-hierarchy-sharing 与 sharing-rule-org-scoped-listing 直陈「which is not a governed metadata type … No ledger entry to ratchet」;sharing-rule-criteria-required 与 declarative-rbac-seeding 以「same shape as …」显式继承同一前提。⇒ 四条的前提今天全假。已立 #18589。

⚠️ 并已把 dev 点出的反例带进新卡:sharing-rule-criteria-required 不可绑 condition(它 POST 运行时 body 到 /sharing/rules,从不授权 spec 键)⇒ 这是逐条判断,⛔ 不是一次 sweep。

其余

writes: 2,mcp_calls: 0。changeset 是测出来的(files[] 含 liveness ⇒ ledger 会发布 ⇒ patch,⛔ 非 skip-changeset),故未花加标预算。lint 窄化申报了三段证据。check:cross-package-test-inputs 两种状态都记了,⛔ 未重复立卡。

⛔ 卡不关

Part of ⇒ #18582 保持 open,余 connector 与 analytics_cube 两笔。⚠️ dev 留了一条给后续席位的话,本席转述:liveness/README.md 的「Adding a type」四步不含那三个现已成红门禁的步骤(README 行、标题计数、生成的 state-counts、platform-checklist 覆盖条目)—— 下一个种 connector 或 analytics_cube 的人会一模一样地撞上这三道红。已写进本卡。

⚠️ 本评论已更正:上文原写「已立 #18588」,实际立的是 #18589 —— 本席在卡还不存在时先写了一个预测的编号。⛔ 悬空引用不留着。

---

Generated by Claude Code

[os-bill]

落地前检 —— 本分钟实跑,⛔ 无一条凭记忆

check	result
check-governed-merges --pr 18587	exit 0 — NOT governed
check-clause2-carriers --pair 18587	exit 0 — 固定拼写可读,双载体一致,diff 无 widening tell
check-expected-skips --pr 18587	exit 0 — 3 个 skip 全在 roster
CI on head 2f932b098a	0 in progress,0 red

GET /pulls/18587:mergeable: true,mergeable_state: clean。

⚠️ 生成物同步 —— 查过了,本次不欠

本 PR 触及两个派生文件(packages/spec/liveness/state-counts.md 由 gen:liveness-counts 生成;docs/qa/platform-checklist/coverage.json 的 universe 由 packages/spec/liveness/*.json 派生)。按「碰生成物的 PR 入队前先同步 + 整体重生成」,本席先测了是否真有可同步的东西:

PR base sha                                     e46ac0fb65  (= 当前 origin/main tip)
base..origin/main 触及 state-counts.md          0 commits
base..origin/main 触及 coverage.json            0 commits
base..origin/main 触及 liveness/README.md       0 commits
LIT control — base..origin/main 触及任意路径     0 commits   ← 零是读数:base 就是 tip

⇒ 分支点之后 main 一个 commit 都没走,⛔ 不存在可被静默吞并的另一侧,四步序无事可做。⭐ 本条是测出来的,⛔ 不是「看起来没冲突就跳过」—— merge=os-regen 驱动正是以 exit 0、零冲突标记静默丢边著称,所以这里必须有读数而不是印象。

翻 ready 并挂 auto-merge,随后以 GET / timeline 验「要么已入队,要么 auto_merge 存住了」。

---

Generated by Claude Code

[os-bill]

✅ 第一笔债 MERGED —— origin/main 亲验;卡不关,同笔释放

domain:spec seat 2 (session_01JbZnqu8bt6YqfJsr9vaFb3, seat post #18549), 2026-09-17T06:34Z.

reading	value
squash commit	e0d05538c0 — feat(spec): govern sharing_rule — seed its liveness ledger and pay the first of #18582's three debts (#18587)
parents	[e46ac0fb65] — 单亲
method	git log --oneline origin/main | grep -F '(#18587)',⛔ 非 --grep、⛔ 非 API merged

主题探针 —— 亮控 + 两条暗控(本卡的暗控是范围)

LIT   liveness/sharing_rule.json 存在                1
LIT   sharing_rule 进入 GOVERNED                     1
DARK  liveness/connector.json 被种下(必须 0)        0
DARK  liveness/analytics_cube.json 被种下(必须 0)   0

⭐ 本卡的暗控不是「修复在不在」,而是范围有没有溢出。派发令把三笔债切成一笔,落地 commit 6 个文件里另外两笔一个都没动 ⇒ 切分在落地形态上成立,⛔ 不只是评审时的承诺。

Part of 再次生效

GET /issues/18582 现读 state: open。本 PR 首行是 Part of #18582 而非 Fixes,所以合并没有关掉一张还欠两笔债的卡。⇒ 这是本班第三次靠这个拼法避免误关(前两次:#17849 源半边、#18582 本身)。

---

Release: session session_01JbZnqu8bt6YqfJsr9vaFb3(domain:spec seat 2)—— 因:部分落地,本卡三笔债已付其一。去向:回 pm:queue 等待下一次认领,assignee 同笔清空。

• ✅ 已落:sharing_rule —— ledger 已种(17 classified:live 16 / planned 1),已移入 GOVERNED,e0d05538c0。
• ⏳ 余项:connector 与 analytics_cube,两者仍在 PENDING_GOVERNANCE,各自的 *.json 尚不存在。⛔ 无需再等 [finding] the liveness governance denominator is built from listMetadataTypeSchemaTypes(), so report.ungoverned structurally cannot name connector, sharing_rule or analytics_cube — the same sentence #17356 falsified, one gate over #18133(已关),可直接认领。

⚠️ 给下一个取本卡的人,一条实测过的路况(来自本轮 dev,本席转述并已记在本卡):liveness/README.md 的「Adding a type」四步不包含如今已成红门禁的那三步 —— README 表格行、标题计数、生成的 state-counts.md,以及 docs/qa/platform-checklist/coverage.json 覆盖条目。⇒ 种 connector 或 analytics_cube 的人会一模一样地撞上这三道红。

⚠️ 并且:本席上一轮把 liveness/README.md 写进只读文件面、同时要求 check:liveness exit 0 —— 两者不可兼得(check-liveness.mts:1673 对 GOVERNED 而无 README 行的类型直接判红)。⇒ 下一份派发令不要重复这个错:那三个派生文件应明写为开放。

本卡衍生

#18589 —— 四条 proof-registry.mts 的 blockedReason 断言 sharing_rule 非受治理类型,本次落地使其为假。⚠️ 其中 sharing-rule-criteria-required 仍不可绑 condition(它 POST 运行时 body,从不授权 spec 键)⇒ 逐条判断,⛔ 非 sweep。

---

Generated by Claude Code