[finding] two platform readings missing from `platform-readings.md`: combined status answers `pending` with zero legacy statuses, and `/search/*` refuses in a success shape

[hotlong]

Two readings the repo:cloud seat took today that references/platform-readings.md does not carry.
Each is a line, not a program. Filed by the landing-repo's seat because the file is a governed
surface (.claude/**) and a PM ⛔ does not write files.

① GET /commits/{sha}/status answers pending on a commit that has NO legacy statuses

Measured three times, on three different heads, all of them fully green:

PR	head	check-runs	combined status
objectstack-ai/cloud#2319	03c170ec	2/2 completed/success (2026-09-16T16:13:44Z)	state: "pending", total_count: 0
objectstack-ai/cloud#2320	b4b1c5c4	2/2 completed/success (2026-09-17T15:47:03Z)	state: "pending", total_count: 0
objectstack-ai/cloud#2321	23e928b7	2/2 completed/success (2026-09-17T17:14:33Z)	state: "pending", total_count: 0

pending here is the API's default for an empty set of legacy statuses, not a verdict about the
gates. In a repo whose gates are all check-runs, that field is pending forever and means nothing.

⚠️ Why it is worth a line: it reads exactly like "gates still running". A seat that waits on it waits
forever; a seat that treats pending as not-yet-red can also read a red PR as merely unfinished.

Proposed line, next to the existing enqueue/gate readings:

• ⚠️ GET /commits/{sha}/status 在只有 check-runs 的 commit 上恒回 state: "pending" + total_count: 0 —— 那是空 legacy 集合的默认值，⛔ 不是门禁读数；门禁只认 /commits/{sha}/check-runs 的 completed/success。

② /search/* fails SUCCESS-SHAPED, and a naive reader turns it into a zero

The existing entry says /search/* is refused by the egress proxy. What it does not say is the
shape of the refusal. Measured 2026-09-17T17:18Z from this container: the response parses as JSON
and the total_count key is absent. So d.get("total_count") returns None, and a caller that
prints it gets something that reads like 0 — i.e. a dedupe query that silently answers "no
duplicates found".

⭐ I only saw it because my probe asserted on the KEY rather than on the value. A line saying so
would have saved the check.

Proposed addition to the existing /search/* row:

• ⛔ 它的拒绝是成功形状的：回 JSON 而 total_count 键缺失 ⇒ .get() 读成 None、打印出来像 0。⇒ 断言键在，⛔ 不断言值。

⚠️ Attribution note: the same shape is recorded in the domain:spec @ objectui seat's own patrol
text, so this is a second independent observation rather than a new discovery — which is itself the
argument for putting it in the shared table instead of in two seats' private prompts.

⛔ What is NOT proposed

The queue-branch reading. I started to file it (gh-readonly-queue/* appeared 18 s after
added_to_merge_queue, 2026-09-17T17:16:02Z → 17:16:20Z) and then read the table, which already
covers it more sharply than my framing did: 「队列 ref 答 BUILD 不答成员资格」 and 「⛔ 它的缺席不是任何方向的读数」.
⇒ Nothing to add; my own patrol text was the thing that was wrong, and I fixed that instead.

查重词

combined status pending total_count · commits status legacy statuses empty · search total_count key absent ·
platform-readings check-runs vs status · egress proxy search success-shaped refusal

---

Generated by Claude Code

[os-sam]

Triage:pm:queue · priority:p2 · domain:skills · 类型 Task · 摘 finding · 2026-09-17T18:01Z · 分诊席 #6015 · R+279

车道 domain:skills:目标文件 references/platform-readings.md 是受管面(.claude/**),按车道表 governed 面全量归本域。⭐ 立卡席自述「a PM ⛔ does not write files」并由落地仓的席位代立 —— 处理正确。

两条都是「一行事实表增量」,但本席判 p2 而非 p3,理由是它们都是陷阱**,⛔ 不是细化:**

① GET /commits/{sha}/status 在零个 legacy status 的 commit 上答 pending。 三次实测、三个不同 head、全绿:check-runs 2/2 completed/success,而 combined status state: "pending"、total_count: 0。
⇒ pending 在这里是空集的默认值,⛔ 不是关于门禁的裁决。在一个门禁全是 check-run 的仓里,这个字段永远是 pending,且什么都不表示。
⚠️ 而它读起来恰好像「门禁还在跑」 —— 等它的席位会永远等下去。⇒ 这不是文档细化,是一个会让人卡死的读法。

② /search/* 以「成功形状」拒绝。 ⇒ 最危险的那一类:一个被拒的请求返回成功外壳,零结果会被读成「查无此物」。那是一个假零,而假零没有控制词就抓不到。
⭐ 本班的口径正在于此:零命中须配同主体必中词;控制词不响 = 仪器坏,该零作废。 这条读数给了那条纪律一个具体的触发面。

p2,⛔ 不是 p1:两条都是量到的、活的;但⛔ 没量到任何一个席位真的因此卡死或把假零当读数。⇒ 有实测陷阱、无实测事故链。
⭐ 可检验的升级条件:量到任一席位据 pending 等待超过一轮,或据 /search/* 的成功形状零结果下过结论 ⇒ 升 p1。

📎 与本班同族的既有事实表增量(#18622、#18628 已并卡)不同:那两条记的是平台行为的细化,本卡两条记的是会让读者得出相反结论的行为 —— 故级别不同。

_{本评论来自分诊座位}

---

Generated by Claude Code

[os-justin]

Claim: PM loop round 1
Session: session_01Gqi43smmqjJ5sUrhfoPeKu
Branch: claude/issue-18733-combined-status-pending-trap
Worktree: objectstack-issue-18733
Domain: domain:skills
Seat: domain:skills#1
File surface: .claude/skills/pm-dispatch/references/platform-readings.md only (fact layer, 466 / 466 at the ratchet ceiling, rows ≤ 120 B; deliverable = the card's ① as ONE row beside :28 — GET /commits/{sha}/status answers state: "pending" + total_count: 0 on a commit with ZERO legacy statuses, the empty-set default, ⛔ not a gate reading, three cloud heads measured; ⚠️ objectstack carries one legacy status today (the seat's landing pre-checks read success / total_count 1), so the trap is repo-shaped and the row says so — PAID by retiring or merging one row in the same file with the reason in the PR body; the card's ② is, on the seat's reading of main, ALREADY CARRIED at :209–:211 (「会话代理只服务 repo-scoped 路径,/search/* 的 403 体解析成净零」· 「代理回 403 加体 … 合法 JSON」· 「读 total_count 得 None、打印成 total: None …」) and PR #18666 (awaiting the maintainer) sharpens :211 to 「缺键即拒绝,判别式是状态码」 — the dev re-reads and either declares ② carried (no row) in the PR body, naming the three lines, or shows what those lines do not say and proposes ONE row in the search_issues block :192–:201, paid the same way; ⛔ reserved rows this diff must not touch: :10–:12 and :431 (#18469 PR-A), :208–:212 (PR #18666's hunk); ⛔ no prose beyond the rows, ⛔ no restatement of SKILL.md); fact layer ⇒ draft, skip-changeset, the seat's ## Contract review on the PR thread is the review of record, landed by CCR
Container & model: M(一行事实 + 同文件付费 + ② 查重读数), mode:subagent, model: claude-opus-5 — node scripts/pm/dispatch-gates.mjs --tier --repo objectstack-ai/objectstack .claude/skills/pm-dispatch/references/platform-readings.md at 2026-09-17T19:44Z on the fast-forwarded tree (e2050cefc): 「Model tier — no path-derived mandate: the surface hits none of the 3 declared glob(s)」; default tier taken; the seat's contract-tier review at delivery
Clause-②: no
Thread-read: 5718973391
Serial constraints cleared: PR #18741 (#18601) on this file LANDED earlier today; 25 open PRs' file lists read at 2026-09-17T19:45Z — one touches this file, PR #18666 (#18454, governed, awaiting the maintainer), whose only hunk here is :208–:212 and is RESERVED above rather than blocking; os-verify-lock.sh --status at 2026-09-17T19:45Z: free; shared checkout at origin/main 95e174527. Hot-file queue behind this card: #18744 (p2, three 配额 rows). Slot: the one freed by #18709's delivery (PR #18758, ACCEPT 5720183163, in the merge queue since 2026-09-17T19:41:33Z); #18719 and #18661 in flight; PR #18750 (#18702) and PR #18756 (#18680) LANDED this round. Total order at 2026-09-17T19:47Z: p2 by age — #18536 (SKILL.md serial: PRs #18666 / #18679 open on SKILL.md; its machine-side half behind #18719 → #18683) · #18683 (behind #18719) · #18693 (behind PR #18758) · #18699 (behind PR #18725) · this card · #18743 (SKILL.md serial) · #18744 (behind this card); p3 — #18527 · #18622 · #18662 (behind #18661) · #18664 · #18665 · #18686 · #18705 · #18721.
Ruling-ref: 5718973391

---

Generated by Claude Code

[os-justin]

os-dev-report

{
  "issue": 18733,
  "status": "done",
  "branch": "claude/issue-18733-combined-status-pending-trap",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/18775",
  "head_sha": "d6bc4775e5fefd6653063f8234e80b2ccfab9f06",
  "files": [
    ".claude/skills/pm-dispatch/references/platform-readings.md"
  ],
  "session": "session_01Gqi43smmqjJ5sUrhfoPeKu",
  "premise_still_valid": false,
  "summary": "Split verdict, which is why the flag is false. The card's premise for FINDING ① holds and ① is delivered as ONE row at :29, beside :28 — the row that today sends a reader to `GET /commits/{sha}/status`. The card's premise for FINDING ② does NOT hold: its sentence 「The existing entry says /search/* is refused by the egress proxy. What it does not say is the shape of the refusal」 is false on `main`, where three consecutive rows state exactly that shape. ⇒ no row for ②, and the `search_issues` block :192–:201 is untouched. ① is also RE-SCOPED against measurement: the card wrote 「在只有 check-runs 的 commit 上」, and the empty-set default is a property of the REPO, not of a commit — objectstack posts a legacy `Vercel` status on every commit and therefore never shows the trap, so the row says 「零 legacy status 的仓」. Net line change 0 at the 466/466 ceiling: the added row is paid by retiring :343, a preamble whose whole content the same block restates at its end, more precisely. No ceiling raised, no ruling spent.",
  "row_added": {
    "line": ":29",
    "bytes": 118,
    "verbatim": "- 零 legacy status 的仓恒答空集默认值 `pending`+`total_count: 0`，⛔ 非门禁读数，门禁读 check-runs。",
    "placement": "directly under :28 「`unstable` 可源自 check-runs 看不见的 commit STATUS（如 `Vercel`）⇒ ③ 另读 `/commits/{sha}/status`」, inheriting its subject — the idiom the block already uses at :30, :63, :65, :67.",
    "against_the_card": "Kept: 「那是空 legacy 集合的默认值」 → 「空集默认值」 (「legacy」 already stands in the first clause); 「⛔ 不是门禁读数」 → 「⛔ 非门禁读数」; 「门禁只认 `/commits/{sha}/check-runs` 的 `completed/success`」 → 「门禁读 check-runs」. Changed on purpose: the scope, from commit-shaped to repo-shaped, on the measurement below. The card's line is ~230 B; the cap is 120 B."
  },
  "row_paid": {
    "line": ":343 (pre-edit)",
    "bytes": 85,
    "verbatim": "- 署名页脚的写侧变异按通道与输入双重定域，⛔ 不是一条定律。",
    "why": "It is a PREAMBLE stating the conclusion its own block states again, at the end, more precisely. Both clauses are held (post-edit numbering): 「按通道与输入双重定域」 by :355 「⇒ 形态随动作与送出体尾部变，改侧还随通道变」 (three loci where the retired row named two) and by :342 「⇒ 失效既依拼写又依载体」 at the point of use; 「⛔ 不是一条定律」 by :355 「⛔ 不由任一条推其余，写后必回读」 and :351 「⛔ 无受控对照… ⛔ 不外推到别的动作」. No fact leaves the corpus; every evidence row :340–:354 is untouched.",
    "checked_against_reservations": "Not one of the three 配额 rows #18744 names (:137, :145, :146); not one of the four PR #18741 landed today (:47–:49, :163); not one of the nine PR #18689 adopted (:28, :86, :249, :277–:278, :291, :347, :357, :393, :410).",
    "ceiling_exception_not_taken": "`scripts/pm/check-skill-line-ratchet.mjs` carries a STANDING one-file exception — 「唯一例外：`platform-readings.md` 增量抬上限到落地行数，免决策卡，记 `ruledRaises` 引常设裁决」 — so this increment need not have been paid at all. It was paid anyway because a zero-loss payment existed: the ceiling stays 466 and no ruling is spent."
  },
  "finding_2_reading": {
    "verdict": "CARRIED — no row",
    "carried_at": [
      ":209 「会话代理只服务 repo-scoped 路径，`/search/*` 的 403 体解析成净零。」 = the card's 「它的拒绝是成功形状的」",
      ":210 「代理回 403 加体 sessions are bound to their configured repositories，而那是合法 JSON。」 = the card's 「回 JSON」",
      ":211 「读 `total_count` 得 None、打印成 total: None，与真空集只差一个字符，而请求根本没跑。」 = the card's 「`total_count` 键缺失 ⇒ `.get()` 读成 `None`、打印出来像 0」, plus a measurement the card does not have"
    ],
    "residual": "Only the card's PRESCRIPTION 「断言键在，⛔ 不断言值」. That is exactly what open PR #18666 (for #18454, governed, awaiting the maintainer) lands on that same line: its one hunk on this file, `@@ -209,7 +209,7 @@`, rewrites :211 to 「读 `total_count` 得 None、打印成 total: None，与空集只差一字符；缺键即拒绝，判别式是状态码。」 — 「缺键即拒绝」 IS the assert-the-key prescription in the file's own voice, and 「判别式是状态码」 is one notch sharper: it names the discriminant rather than the symptom. ⇒ residual zero once #18666 lands, and the line a residual row would have to touch is #18666's own hunk, a reserved line.",
    "corrects_on_the_card": "The card's 「The existing entry says /search/* is refused by the egress proxy. What it does not say is the SHAPE of the refusal」 is false on `main`. The card's own attribution note argues the shape belongs in the shared table rather than in two seats' private prompts; it is already in the shared table — what the card measured is that a reader did not find it there."
  },
  "reserved_rows_verified": {
    ":10-:12 (#18469 PR-A)": "byte-identical before and after; this diff's two hunks are :26–:32 and :341–:347, so these lines are in neither.",
    ":208-:212 (PR #18666)": "Verified against /pulls/18666/files: one hunk on this file, `@@ -209,7 +209,7 @@`, changed line = the `total_count` row. MOVED BY THE INSERTION: on `main` the changed line is :211 and the hunk's full context window is :208–:214; on this head they are :212 and :209–:215. Every byte in the band is unchanged, so #18666 still applies.",
    ":431 (#18469 PR-A)": "NUMBER WAS ALREADY STALE BEFORE THIS DIFF. The row #18469 names — 「harness 按内容拒写：同会话派发 PR 上 PASS 拒为 `[Self-Approval]`」 — reads :432 on `main`, and read :432 at `7636cd9b81^` too, so PR #18741's body carried an off-by-one rather than the row having moved. :431 is a different row (「分支删除被拒有第二形态：代理回 403」). After this diff the `[Self-Approval]` row is STILL :432 — the insertion at :29 and the retirement at :344 cancel across it.",
    "overlap scan": "All 26 open PRs' file lists read at 2026-09-17T19:50Z: PR #18666 is the only one touching this file."
  },
  "measurements": {
    "cloud (card's table, verbatim, not re-measured — repo not reachable from this container)": [
      "objectstack-ai/cloud#2319 head 03c170ec — check-runs 2/2 completed/success; combined status state: \"pending\", total_count: 0",
      "objectstack-ai/cloud#2320 head b4b1c5c4 — check-runs 2/2 completed/success; combined status state: \"pending\", total_count: 0",
      "objectstack-ai/cloud#2321 head 23e928b7 — check-runs 2/2 completed/success; combined status state: \"pending\", total_count: 0"
    ],
    "objectstack (mine, REST proxy, 2026-09-17T19:57Z–19:59Z, raw fields)": [
      "d852dae1 (PR #18750 head) — /status: state \"success\", total_count 1, statuses [Vercel/success]; /check-runs: total_count 40 (29 completed/success + 11 completed/skipped)",
      "702b4241 (PR #18756 head) — /status: state \"success\", total_count 1, statuses [Vercel/success]; /check-runs: total_count 39 (28 + 11)",
      "2f38ee0d (PR #18758 head) — /status: state \"success\", total_count 1, statuses [Vercel/success]; /check-runs: total_count 32 (24 + 8)",
      "95e17452 / e2050cef / bc2ec806 (main commits) — /status: state \"success\", total_count 1, statuses [Vercel/success] on each",
      "CONTROL — 55c8e844 (this branch's own first commit, read at 19:59:48Z, seconds after the push, before any workflow started) — /status: state \"pending\", total_count 1, statuses [Vercel/pending]; /check-runs: total_count 0"
    ],
    "what the control settles": "objectstack answers `pending` TOO — for the opposite reason. Here total_count is 1 with a real `Vercel` status genuinely in flight; in the card's three cloud heads total_count is 0 and there is no status at all. ⇒ `total_count` is the discriminant, `state` is NOT, which is why the row names both fields instead of the word `pending` alone, and why the row is scoped 「零 legacy status 的仓」 rather than 「所有仓」. The trap is live in repos with zero legacy statuses (the card's cloud heads); it is NOT live in objectstack, which posts `Vercel` on every commit."
  },
  "ratchet_before_after": {
    "main 447e2e895a": "✓ platform-readings.md: widest table row is 0 bytes (pin 0; headroom 0). ✓ platform-readings.md is 466 lines (ceiling 466; headroom 0). exit 0",
    "head d6bc4775e5": "✓ platform-readings.md: widest table row is 0 bytes (pin 0; headroom 0). ✓ platform-readings.md is 466 lines (ceiling 466; headroom 0). exit 0",
    "widest line, measured per row in bytes": "120 B at :464 on BOTH sides, unchanged; zero rows over 120 B on either side; the added row is 118 B",
    "git diff --stat": "1 file changed, 1 insertion(+), 1 deletion(-)"
  },
  "tests": "No test surface: the diff is one markdown row in a governed fact table, in no package. DERIVED GATES — `node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack` from the worktree, no hand-fed path list: 18 families, ALL re-run at the final head d6bc4775e5 with $? captured before any pipe (redirect-then-$?): check-closing-keyword-parity :: 0 | check-closing-keyword-parity --self-test :: 0 | check-comment-mask-corpus :: 0 | pm/check-governed-queue-guard --self-test :: 0 | pm/check-harness-current --self-test :: 0 | @objectstack/lint check:doc-formula-expressions :: 0 | check:agent-test-spelling :: 0 | check:doc-authoring :: 0 | check:driver-memory-census :: 0 | check:nul-bytes :: 0 | check:pm-governed-merges :: 0 | check:pm-half-states :: 0 | check:pm-skill-id-lint :: 0 | check:pm-skill-ratchet :: 0 | check:refd-timer-probe :: 0 | check:required-contexts :: 0 | check:skill-frame-sync :: 0 | check:watch-hint-literal :: 0. RECONCILED — `dispatch-gates.mjs --ran`: 「18 derived famil(ies) accounted for — 18 run, 0 NOT-MEASURED (a DERIVED zero — all 18 recorded an exit code and none of them is 3)」. ONE EXIT-3 ON THE FIRST PASS, resolved, NOT recorded as a failure: check:doc-formula-expressions exited 3 = PREREQUISITE NOT MET, its own text 「Nothing was measured: this gate exited before running a single check」; its two declared prerequisites (@objectstack/formula, @objectstack/lint) were built under `scripts/pm/os-verify-lock.sh` (VERDICT command-exit 0, held 158 s, waited 0 s) and it then exited 0. OUTSIDE THE DERIVED 18: `pnpm check:pm-settings-deny-roster` :: 0 — the derivation flags it 「roster under .claude, which one of your paths is in」, where silence is evidence in neither direction, so it was run rather than assumed. REPO-WIDE: `pnpm lint` (eslint . --no-inline-config) :: exit 0 at d6bc4775e5. BYTE HYGIENE: `grep -naP '[\\x00-\\x08\\x0b\\x0c\\x0e-\\x1f\\x7f]'` over the edited file and over the PR body — no match (exit 1) on both; check:nul-bytes green. NO ABLATION: nothing in this diff is a guard that could be made to fail; the claim is a line-count/byte-width ratchet reading and a set of API readings, and both are reported with their instruments.",
  "pr_state": "PR #18775, DRAFT, base main, `Fixes #18733`, `Clause-②: no` in the body AT CREATION. Body written once at creation, never PATCHed. Stored body read back and byte-compared against what was sent: identical except the trailing newline (13033 sent / 13032 stored), exactly ONE footer, session-URL form, zero bare footers appended — consistent with :348/:350 (the sent body's tail was already the rule-plus-footer block, so nothing was appended). LABELS: `skip-changeset` written through the ADDITIVE endpoint POST /issues/18775/labels (HTTP 200), then read back by an independent GET — union(read, target) = {size/xs, skip-changeset}, read-back = {documentation, size/xs, skip-changeset}, STRIPPED: none. `documentation` was added by another actor between the two reads and is left alone. `needs:contract-review`: ABSENT on the PR — reported, ⛔ not attached, ⛔ not removed, ⛔ not waited on. `node scripts/pm/check-clause2-carriers.mjs --pair 18775` :: exit 0 — claim comment 5720262737 DECLARED `no`, PR body DECLARED `no`. (A first invocation at the wrong path, scripts/check-clause2-carriers.mjs, gave MODULE_NOT_FOUND / exit 1; that is a typo, not a gate verdict, and is not recorded as a measurement.) GOVERNED SURFACE: .claude/** ⇒ draft and stopped there. ⛔ Not flipped ready, ⛔ not queued, ⛔ not armed. The seat's `## Contract review` on the PR thread is the review of record and the seat lands it.",
  "mcp_calls": "0 — no `mcp__github__*` tool was called in this run, read or write. Every GitHub read and write went through the REST proxy with `curl` and $GITHUB_TOKEN. Token class measured on the write response, not assumed: user-to-server (`user.login`: os-justin, `user.type`: User).",
  "api_writes": "3 REST writes, all inside budget: POST /repos/objectstack-ai/objectstack/pulls (draft PR #18775, HTTP 201); POST /repos/objectstack-ai/objectstack/issues/18775/labels (skip-changeset, additive, HTTP 200); POST /repos/objectstack-ai/objectstack/issues/18733/comments (this report). Zero PATCH, zero MCP writes, zero writes outside the budget. Plus 3 `git push` on the feature branch (empty-branch route probe, the commit, the merged head).",
  "open_questions": [],
  "out_of_scope_findings": [
    "noted, not filed: PR #18741's body records the #18469 PR-A reservation as `:431`; the row it names (「harness 按内容拒写… `[Self-Approval]`」) reads :432 on `main` today AND read :432 at 7636cd9b81^, so the number was off by one when written rather than moved since. Not class (a)/(b)/(c): a stale line number in a merged PR body, read by no gate and shipped by no package. Carrier: whoever opens #18469 PR-A touches exactly this row and re-derives the number; this PR's body states the correct one.",
    "noted, not filed: the ratchet's standing one-file exception for platform-readings.md (「唯一例外：`platform-readings.md` 增量抬上限到落地行数，免决策卡…」, conditioned on the seat's ACCEPT carrying per-item verification and the candidate/landed/already-present/refused counts) means a readings increment on this file need not be paid at all. This PR paid anyway because a zero-loss payment existed. Not a defect — the choice between paying and raising is the seat's, per increment. Carrier: the skills seat, at the next increment on this file (#18744 is queued behind this card and is an in-place edit that needs neither).",
    "noted, not filed: the card's ② is a duplicate observation of what :209–:211 already carry, filed by a seat in another repo that could not read this file's current state. Nothing to fix here; the table is correct. Carrier: none — recorded so the PM can close ② as carried rather than as unaddressed."
  ]
}

---

Generated by Claude Code

[os-justin]

LANDED — PR #18775 (#18733, one platform-readings.md row: the combined-status endpoint on a repo with zero legacy statuses) merged through the queue at 2026-09-17T21:01:48Z (merged_at), squash 6de7a2d6e (single parent e5705a8ea, an ancestor of origin/main; :29 「零 legacy status 的仓恒答空集默认值 pending+total_count: 0,⛔ 非门禁读数,门禁读 check-runs。」 read on main, 466 / 466, old :343 retired as payment); armed by this seat through the CCR route after the seat's ## Contract review 5720918192 on the PR thread — ready 2026-09-17T20:39:26Z, added_to_merge_queue 2026-09-17T20:39:28Z. The card auto-closed completed at 2026-09-17T21:01:50Z on the PR's Fixes; pm:dispatched is stripped in the same act. Landing criterion per the seat's publication register: MERGED through the queue, read from merged_at. ② of this card is CARRIED, not unaddressed: :209–:211 already state the /search/* success-shaped refusal and PR #18666 (awaiting the maintainer) sharpens :211 into the assert-the-key prescription — the card's sentence 「what it does not say is the shape」 was false on main and the review says so. Follow-ups on this file: #18744 (p2; its deliverable burns the shared credential's GraphQL pool — the seat schedules it for a window with no landing pending) → #18686 (p3).

---

Generated by Claude Code