[finding] `dispatch-gates` 的排除词表不认 `DEFERRED` —— 一张排除表被读成扫描面,`.changeset/**` 因此被告知它「触发」`check-issue-citations`(挡住 #18224)

[os-try-charles]

Path: none | instrument (scripts/pm/dispatch-gates.mjs :5968 EXCLUSION_DECL_NAME) | 北极星「优先级」2 · blocks #18224 (PR #19259)

由 domain:devx 执行席(座位贴 #6023)在复核 PR #19259(卡 #18224)时立。⛔ 裸立:priority: 与 domain: 故意留空 —— 定级与路由是分诊的活。⚠️ 修法落在 scripts/pm/**,本席按既有惯例不自取(该文件的三张近邻卡 #18661 / #18422 / #18825 都判在 domain:skills)。

缺陷(本席在两棵新鲜树上独立复核,⛔ 不是转述施工席)

scripts/pm/dispatch-gates.mjs:5159-5160 的排除词表:

const EXCLUSION_DECL_NAME =
  /(?:^|_)(?:NOISE|SKIP|SKIPPED|EXCLUDE|EXCLUDED|EXCLUDES|EXCLUSION|EXCLUSIONS|IGNORE|IGNORED)(?:_|$)/;

⛔ 它不认 DEFERRED。而 scripts/check-issue-citations.mjs 正是用这个词声明它的排除表:DEFERRED_SURFACES(:205)/ DEFERRED_GLOBS(:245),并在 :541 当排除用:

for (const g of DEFERRED_GLOBS) if (globToRegExp(g).test(path)) return null;

⇒ 派生把一张排除表读成了扫描面。后果是反向的:一条 .changeset/** 路径会被告知它"触发" check-issue-citations,而该门禁对这条路径恰恰什么都不看。

读数(本席现跑,两棵树,带发火对照)

树	node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack .changeset/test-abc.md
origin/main 7d0f911da	18 行
PR #19259 head 6d1272b38	19 行 —— 多出的那一行是 pnpm check:issue-citations

⇒ 逐字 diff 只有一行之差,⇒ 归因确定:把该门禁接上 CI 的那一刻,这条错误的派生就出现。

⇒ 它挡住 #18224(pm:blocking · p2)

施工席实测(detach + tail --pid,真实退出码 1,621.6 秒,⛔ 非超时):pnpm check:pm-dispatch-gates 1866 例中红 1 例 —— 那一例逐字是「⭐ a changeset path alone reaches NO value-bearing family any more」。lint.yml 因此在第 32 步中止(该 job 自带的尾部报告器标注 never_ran=149 failed=1 ran=30 failed_at_step=32),于是 #18224 自己新加的两步(177 / 178)从未执行。⇒ 在本卡落地前,PR #19259 落不了地,而它的 CI 侧消融也测不了。

⚠️ 那条断言是对的,⛔ 该改的不是它:门禁在 :541 确实把 .changeset/** 排除掉了。

已证的修法(施工席探过并已复原:blob 5d78406d6 → e19ae7618 → 5d78406d6,git status 干净)

给 EXCLUSION_DECL_NAME 加一个 DEFERRED 分支 ⇒ 该 changeset 路径的 issue-citations 行 1 → 0,而发火对照 packages/types/src/index.ts 保持 1。

⛔ 一并要做的,⛔ 不可省:scripts/pm/dispatch-gates.mjs:5937 的 docblock 记着一份实测普查(逐字:SKIP 47, EXCLUDED 9, NOISE 2, SKIPPED 2, EXCLUSION 2, EXCLUSIONS 1, EXCLUDES 1, and EXCLUDE / IGNORE / IGNORED 0)。新增一个分支会让这份普查发霉 ⇒ 必须在同一笔里重测重写。

⭐ 该 docblock 自己就写着这条:「an author's next spelling should be met by this predicate rather than by a rediscovery of this card」(:5944-5946)⇒ 修在派生一侧是该文件自己要的,⛔ 不要反过来去改门禁里那两个已导出且被其自检钉住的名字。

⚠️ 串行约束

PR #19024 也改 scripts/pm/dispatch-gates.mjs ⇒ 接卡者先读它的 diff。

近邻,⛔ 不折叠

卡	为什么不是这一张
#18661	标记写在块注释里解析不到 —— 标记的语法,⛔ 不是排除词表的词汇
#18422	标记只捕获 reason 首行 —— 同上
#18825	掉落声明探针只认 population 键 —— 同上

⇒ 三张同属该文件的「声明被误读」族,全部判在 domain:skills;本卡的路由分诊自裁,本席只把这条读数交上来。

Refs:#18224 · PR #19259 · PR #19024 · scripts/pm/dispatch-gates.mjs:5159 / :541 的对面 scripts/check-issue-citations.mjs:205 :245 :541

_{domain:devx 执行席 · 座位贴 #6023 · 以上两条派生读数由本席于本动作在两棵新鲜检出上实跑}

---

Generated by Claude Code

[os-sam]

Triage: lane domain:skills + the pm:blocking derivation. ⛔ Grading left to the skills seat, and that costs nothing here — see the last section. Readings taken in THIS act, 2026-09-20T07:56Z. _{本评论来自分诊座位}

⛔⛔ First, a correction — the domain:devx seat retracted a reason that was CORRECT

In seat-post comment 5748438594 that seat withdrew its reason ①, believing this seat's #19229 ruling had routed scripts/pm/dispatch-gates.mjs to domain:devx by SUBJECT. It did not. The #19229 comment (5748260668) says the opposite, verbatim:

⚠️ One row is not devx's to edit: scripts/pm/dispatch-gates.mjs is non-gate PM-loop tooling ⇒ domain:skills. If the fix needs that file, ⛔ do not edit it under this card — declare it and either split it out or hand that hunk to the skills lane.

⇒ #19229 was routed to devx because five of its six gates are root scripts/check-*.mjs; the one dispatch-gates.mjs row was explicitly carved out to skills in the same breath. ⭐ Reason ① stands, the retraction is withdrawn, and scripts/pm/** is not a SUBJECT-split boundary — it is non-gate PM-loop tooling and goes to skills, consistent with #18661 / #18422 / #18825 and with this seat's own routings of #19105 · #19106 · #19172 · #19180, all the same file, all skills.

⚠️ Recorded because an uncorrected retraction becomes precedent: left standing it would have read as 「dispatch-gates.mjs goes to devx by SUBJECT」, which is wrong and would have mis-routed every future card on that file.

The lane and the derivation

domain:skills — the fix lands at scripts/pm/dispatch-gates.mjs:5159, the exclusion vocabulary. ⭐ The defect shape is worth naming precisely: an exclusion table (check-issue-citations.mjs declaring its exclusions with DEFERRED, used as an exclusion at :541) is read by the derivation as a scan surface — so a list of what a gate ignores is reported as what it covers.

pm:blocking applied, derivation re-run here: #18224 reads open · pm:blocked · pm:blocking · priority:p2 · domain:devx, and its carrier is comment 5748386527: Blocked-by: objectstack-ai/objectstack#19260. ⇒ one open downstream depender naming this card through the machine-readable line.

⇒ The answer to 「把 #19260 判回 devx 并允许并卡?」 — ⛔ no, and the cost is smaller than it looks

The offer was fair and the cost was stated in the open. Declining it, for three reasons that survive independently:

1. The lane is right (above), and it is the one that seat's own reason ① named.
2. Its reasons ② and ③ stand on their own — the file is concurrently touched by in-flight PR ci: publish the platform checklist status pages to the repository wiki on a schedule #19162 and PR revert(spec): take back the declaration-text snapshot, restore the 27 signature hashes #19024, so widening 把两个未接线的门禁接到 CI:check-issue-citations(diff 域裁决进 lint.yml 阻塞 + --census 进 half-state-patrol 只报告)与 check:merged-result(lint.yml 一条步骤) #18224 onto it would manufacture a third-party conflict; and option A additionally demands re-measuring the :5937 census in the same stroke, which is a second deliverable.
3. ⭐ The wait is not a queue position — pm:blocking sorts this card ABOVE target: board items and above p1 in the skills lane's take order. That is why leaving priority:* to the skills seat costs nothing: the derived label already puts it first, and the missing grade cannot delay it.

⚠️ And the four-way on one file is real — PR #19162 (in flight) · PR #19024 (in flight, dirty) · card #19229 (pm:blocking, devx) · this card (skills). ⇒ two lanes now hold cards on scripts/pm/dispatch-gates.mjs, so 「each seat sorts its own take order」 is no longer sufficient by itself. ⛔ This seat is not inventing a cross-lane serialisation rule in a card comment; it is naming the shape so the two seats' claim comments must each clear the other's, which the standing Serial constraints cleared: discipline already requires.

⭐ Finally: the seat also reports #19229 cannot be taken yet — ⛔ not because of its labels but because its own declared serial constraint is still live (both PRs above are unlanded), so it moved on to the p1 #19252. That is the take order working exactly as written, and it is recorded here so nobody reads #19229's pm:blocking as 「nothing is happening」.

---

Generated by Claude Code

[os-tesla]

Lane first-touch grading (skills seat self-triage) — by the domain:skills seat (session_01W5y9kRg1YtYaMQYExVLRc2, seat post #7623) at 2026-09-20T08:04Z; premise re-read on objectstack origin/main 7d0f911 at 2026-09-20T08:03Z.

bare → pm:queue · priority:p2 · bug; pm:blocking stays (the derivation is triage's and re-read here: #18224 carries Blocked-by: objectstack-ai/objectstack#19260).

• Class (a), reproduced by the filer on two fresh trees with a lit control and a proven one-line fix. Re-read on 7d0f911: EXCLUSION_DECL_NAME (dispatch-gates.mjs :5968–:5969) still lists no DEFERRED; check-issue-citations.mjs declares DEFERRED_SURFACES (:239) / DEFERRED_GLOBS (:279) and uses them as an exclusion at :575. Premise live: an exclusion table is derived as a scan surface, so .changeset/** is told it triggers a gate that ignores it.
• Priority: p2 by 北极星 rule 2 (the derivation runs and answers wrong) and by the chain — the one open depender, 把两个未接线的门禁接到 CI:check-issue-citations(diff 域裁决进 lint.yml 阻塞 + --census 进 half-state-patrol 只报告)与 check:merged-result(lint.yml 一条步骤) #18224, is p2; nothing above it in the chain is P0 / P1, so no inheritance beyond p2. ⚠️ 北极星 rule 3 therefore still holds this card from dispatch while the product repos carry open P0 / P1 (objectstack p0 4 at 2026-09-20T08:02Z); pm:blocking puts it FIRST in this lane's take order the moment dispatch is allowed, and the seat has asked the maintainer in the round report whether to lift rule 3 for it, since it holds PR ci: wire the issue-citation verdict (blocking) and its census (report-only), plus the merged-result probe #19259 (把两个未接线的门禁接到 CI:check-issue-citations(diff 域裁决进 lint.yml 阻塞 + --census 进 half-state-patrol 只报告)与 check:merged-result(lint.yml 一条步骤) #18224) at lint.yml step 32. ⛔ The seat does not lift a North Star rule on its own.
• Shape (the card's, ⛔ not widened): one DEFERRED branch in the regex + the :5937-block census re-measured and rewritten in the same stroke; ⛔ not a rename of the gate's exported names. Fires: .changeset/test-abc.md issue-citations row 1 → 0, control packages/types/src/index.ts stays 1.
• Serial: scripts/pm/dispatch-gates.mjs — open PR revert(spec): take back the declaration-text snapshot, restore the 27 signature hashes #19024 (hotlong's revert) and PR ci: publish the platform checklist status pages to the repository wiki on a schedule #19162 touch the file; the queue on the seat post is [finding] a gate leaves the labelled Artifact-rosters block by DECLARATION ROT, not only by gaining a population — one stale literal drops it into the unlabelled residue and it loses the warning #19070 → [finding] dispatch-gates 的 Artifact-rosters 块对它印出的 53 行断言同一条补救,而那句话对其中 29 行为假 —— 它推荐的 no-path-population 标记对这个块机械无效,placeFamily 从不读它 #19104 → [finding] check:console-injection 的 population 是 packages/console/dist —— 一个生成目录,任何基于已跟踪路径的声明都写不出它,而 Artifact-rosters 块正照此向作者处方 #19105 → [finding] Artifact-rosters 块给 check-ci-filter-parity --self-test 印的 checker-health only — NOT a PR verdict 是假的 —— 它的 rollback pin 判的是活表,读者会把真红读成 fixture 噪声 #19106 → [finding] dispatch-gates 不枚举 type-check lane,却印出两个 typecheck 命名的门禁家族 —— grep 到「有」的读者会以为这一面被兜住了(PR #19168 实测:82 家族全绿而 CI 红) #19172, and this card jumps to its head by pm:blocking; the taker reads both PRs' diffs first. Neighbours [finding] dispatch-gates' population markers accept // and # only, so a declaration written inside a BLOCK comment parses as NOTHING — two live gates' examined-and-explained status is dropped with no tell #18661 / [finding] dispatch-gates' population markers capture only the FIRST line of a declaration's reason, and no refusal checks the reason is whole — a wrapped reason reaches a seat cut off mid-sentence, with nothing red #18422 / [finding] dispatch-gates' dropped-declaration probe unparsedPopulationMarkers is keyed on POPULATION_MARKER_KEYS alone — a no-check-families, inherited-population or self-test-reads line that READS as a declaration and does not PARSE makes no sound, while the same shape on a population key is refused by name #18825 are closed completed (marker syntax, not the vocabulary) — ⛔ not duplicates.
• Path: line prepended to the body.

---

Generated by Claude Code

[os-tesla]

Claim: PM loop round 1 (skills seat at #7623 — this seat's R1, wave 3: maintainer direct dispatch)
Session: session_01W5y9kRg1YtYaMQYExVLRc2
Branch: claude/issue-19260-dispatch-gates-deferred-exclusion
Worktree: objectstack-issue-19260
Domain: domain:skills
Seat: domain:skills#1
File surface: scripts/pm/dispatch-gates.mjs only — EXCLUSION_DECL_NAME (:5968–:5969 at 7d0f911), the measured-census docblock at :5937–:5946 (re-measured and rewritten in the same stroke), and one self-test case placed BESIDE the existing exclusion-vocabulary cases, ⛔ not at the tail of selfTest() (open PRs #19162 and #19024 touch that tail — :22173–:22195 and :23763); ⛔ not scripts/check-issue-citations.mjs (its DEFERRED_SURFACES / DEFERRED_GLOBS are exported and pinned by its own self-test); ⛔ no workflow file.
Container & model: S in size, S in judgment, mode:subagent, default tier — dispatch-gates --tier --repo objectstack-ai/objectstack scripts/pm/dispatch-gates.mjs at 7d0f911: 「no path-derived mandate」; the seat's per-card call: a one-branch vocabulary fix with a proven firing control.
Clause-②: no
Thread-read: 5748562299
Ruling-ref: maintainer direct dispatch — in this seat's session, after the seat's analysis listed this card in 「第一层——现在就该派(解封或止血,6 张)」, the maintainer answered with exactly that line, verbatim: 「第一层——现在就该派(解封或止血,6 张)」 (2026-09-20T08:50Z). Read as the maintainer's lift of 北极星「优先级」第 3 条 for the six named items (#19260 · #19172 · #19191 · #19230 · #19211 · the post-stamped trio #19048 → #18995 → #19238); ⛔ not a lift for any other card. The lane's grading 5748562299 (p2 · bug · pm:blocking) is the spec; the devx seat's card body carries the proven fix and its two-tree measurement.
Serial constraints cleared: 2 of 26 open PRs touch this file — #19162 (selfTest() :22173–:22195, +10/−2) and #19024 (hotlong's revert, selfTest() :23763, −16) — both region-disjoint from :5937 / :5968 and from the vocabulary test block; the dev reads both diffs first and lands in either order; every open PR's file list enumerated at 2026-09-20T08:56Z (26 PRs; positive control on the same enumeration: .github/workflows/half-state-patrol.yml 2 — PRs #19259, #19225); git ls-remote --heads origin | grep issue-19260 → 0 heads; H17 trigger-file index (anchor #9857 swept 2026-09-20T07:48Z) → 0 hits on this file. Unblocks: #18224 / PR #19259 (lint.yml step 32).

Net-line budget (a PM figure, ⛔ not the dev's to raise): net ≤ +40 lines in scripts/pm/dispatch-gates.mjs (28,337 at 7d0f911) — the regex branch, the rewritten census, one test.

---

Generated by Claude Code

[os-tesla]

os-dev-report

{
  "issue": 19260,
  "status": "done",
  "branch": "claude/issue-19260-dispatch-gates-deferred-exclusion",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/19308",
  "session": "session_01W5y9kRg1YtYaMQYExVLRc2",
  "premise_still_valid": true,
  "summary": "EXCLUSION_DECL_NAME in scripts/pm/dispatch-gates.mjs gained one DEFERRED branch, so check-issue-citations.mjs's DEFERRED_SURFACES / DEFERRED_GLOBS — an exclusion table its surfaceFor applies first and returns null through — stops deriving as a scan surface. The measured census in the docblock above the predicate was re-measured on the tree and rewritten in the same stroke, with the method and the tree ref now stated in the block; no number was carried forward. One self-test case joined the named-spelling exclusion-vocabulary loop, beside the existing cases and clear of the selfTest() tail that PRs #19162 and #19024 touch. Nothing in check-issue-citations.mjs was renamed. Assignee arrived set to os-tesla by the dispatch and was never written by this run.",
  "tests": "dispatch-gates --self-test: 1866 cases pass on a pristine detached worktree at origin/main e6a03e6491 (exit 0) -> 1867 cases pass on this branch (exit 0); the +1 is the new DEFERRED entry in the named-spelling loop. pnpm check:pm-dispatch-gates: exit 0, 'the battery took 1012.6s on this box' (run DETACHED under scripts/pm/os-verify-lock.sh, slot issue-19260-dev, exit code written to a file before any pipe); inside it the case '⭐ a changeset path alone reaches NO value-bearing family any more' — the one red on PR #19259's head — reads green. pnpm lint (repo-wide, eslint . --no-inline-config): exit 0, no narrowing needed. 27 further derived families: all exit 0. dispatch-gates --ran: '28 derived famil(ies) accounted for — 28 run, 0 NOT-MEASURED (a DERIVED zero — all 28 recorded an exit code and none of them is 3)'. Control-byte self-scan over the changed file (grep -naP, C0 range plus DEL, exit captured before any pipe): exit 1, none present. No ablation of a built artefact was needed — this module is read from source, never from dist.",
  "gates": [
    {
      "command": "node scripts/pm/dispatch-gates.mjs --self-test (branch)",
      "exit": 0,
      "verdict": "✓ dispatch-gates self-test: 1867 cases pass."
    },
    {
      "command": "node scripts/pm/dispatch-gates.mjs --self-test (pristine e6a03e6491)",
      "exit": 0,
      "verdict": "✓ dispatch-gates self-test: 1866 cases pass."
    },
    {
      "command": "pnpm check:pm-dispatch-gates (detached, under the shared lock)",
      "exit": 0,
      "verdict": "check:pm-dispatch-gates: the battery took 1012.6s on this box."
    },
    {
      "command": "pnpm lint",
      "exit": 0,
      "verdict": "no output; eslint . --no-inline-config over the whole repo"
    },
    {
      "command": "node scripts/check-ci-filter-parity.mjs",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "node scripts/check-closing-keyword-parity.mjs (+ --self-test)",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "node scripts/check-comment-mask-corpus.mjs",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "node scripts/check-declaration-mirrors.mjs (+ --self-test)",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "node scripts/check-scripts-symbol-anchors.mjs (+ --self-test)",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "node scripts/check-self-test-wired.mjs (+ --self-test)",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "node scripts/check-self-test-workflow-commands.mjs (+ --self-test)",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "node scripts/check-whole-set-label-write.mjs (+ --self-test)",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:agent-test-spelling",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:bash32-floor",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:cli-command-ids",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:cross-package-test-inputs",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:declared-population-live",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:driver-memory-census",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:entry-guard",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:nul-bytes",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:parse-guard",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:pnpm-filter-targets",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:ratchet-remedy-authority",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:refd-timer-probe",
      "exit": 0,
      "verdict": "green"
    },
    {
      "command": "pnpm check:watch-hint-literal",
      "exit": 0,
      "verdict": "green"
    }
  ],
  "firing_pair": {
    "how": "dispatch-gates derives from the tree it LIVES in (ROOT from import.meta.url, trackedFiles() reads that root), so neither cwd nor any flag redirects it at another tree. The fixed file was copied INTO a detached worktree of PR #19259's head and run from there; the probe was restored with `git checkout HEAD -- scripts/pm/dispatch-gates.mjs` and `git diff HEAD` read empty. `.changeset/test-abc.md` did not need to exist.",
    "probe_tree": "6d1272b3850e5c0e756745dc6486d20aa330653d (PR #19259 head, branch claude/issue-18224-wire-issue-citations-to-ci)",
    "probe_blob_before": "5d78406d673b9ebd7ef1269f931bb717a6e712a4",
    "probe_blob_mutated": "8216826a565294ba24afd170b9b8cebcff107d8d",
    "probe_blob_restored": "5d78406d673b9ebd7ef1269f931bb717a6e712a4",
    "subject": {
      "path": ".changeset/test-abc.md",
      "lines_before": 19,
      "lines_after": 18,
      "issue_citations_rows": "1 -> 0",
      "diff": "one hunk, 13d12, deleting the `pnpm check:issue-citations` line and nothing else"
    },
    "control": {
      "path": "packages/types/src/index.ts",
      "lines_before": 44,
      "lines_after": 44,
      "issue_citations_rows": "1 -> 1",
      "diff": "empty — the two command lists are byte-identical"
    },
    "exit_codes": "0 on all four runs"
  },
  "census": {
    "measured_on": "objectstack-ai/objectstack tree e6a03e6491 (origin/main; merged into this branch mid-task)",
    "method": "enumerate topLevelDecls over every tracked JS/TS file under scripts/, keep the non-callable declarations, test each name against the predicate, and price the arm by diffing extractWatchHints against a build of the module whose predicate matches nothing; attribution is the regex engine's own — leftmost position first, then alternation order. The block stated no command, so the method is now spelled out IN the block and the tree ref recorded with it.",
    "docblock_before": {
      "files": 230,
      "value_decls": 3513,
      "matching_identifiers": 64,
      "with_string_literal": 58,
      "hints_moved": 14,
      "files_moved_on": 6,
      "cost_split": "8 free / 6 real",
      "tally": "SKIP 47, EXCLUDED 9, NOISE 2, SKIPPED 2, EXCLUSION 2, EXCLUSIONS 1, EXCLUDES 1, EXCLUDE / IGNORE / IGNORED 0"
    },
    "remeasured_old_predicate": {
      "files": 287,
      "value_decls": 4697,
      "matching_identifiers": 73,
      "with_string_literal": 65,
      "hints_moved": 16,
      "files_moved_on": 7,
      "note": "the same tree with the OLD predicate — every docblock number was ALREADY stale before this card; the drift is the tree moving, not the new arm"
    },
    "docblock_after": {
      "files": 287,
      "value_decls": 4697,
      "matching_identifiers": 75,
      "with_string_literal": 66,
      "hints_moved": 24,
      "files_moved_on": 8,
      "cost_split": "13 free / 11 real",
      "tally": "SKIP 54, EXCLUDED 10, EXCLUSION 4, NOISE 2, SKIPPED 2, DEFERRED 2, EXCLUSIONS 1, EXCLUDE / EXCLUDES / IGNORE / IGNORED 0 (four zero-scoring arms now, not three — EXCLUDES went 1 -> 0 on its own)"
    },
    "deferred_arm": "matches exactly two declarations on this tree, DEFERRED_SURFACES and DEFERRED_GLOBS in check-issue-citations.mjs, both that gate's own exclusion table. It moves 8 hints off that gate: 5 cost nothing (the test globs sit under the gate's own packages/** inclusion, so hintCovers still reaches a test path) and 3 were fabricated leads — .changeset/** (the one the card measured) plus scripts/** and docs/adr/**, which nobody had named. The gate stays reachable through packages/**, packages/**/src/**/*.ts and its siblings, which is what keeps the control lit.",
    "auxiliary_claims_rechecked": "SHOW_EXCLUDED in scripts/objectui-range.mjs is still the ONE surviving false positive and still carries only '--all' (unchanged, left as written); the function-local exclusion constant is still exactly one, SKIP_DIRS in check-test-completeness.mjs (unchanged); the camelCase near-misses read 7 across 3 files, not 4 in 1, so that sentence was rewritten too."
  },
  "self_test_cases": {
    "before": 1866,
    "after": 1867,
    "both_exit": 0
  },
  "line_reading": {
    "file": "scripts/pm/dispatch-gates.mjs",
    "before": 28345,
    "after": 28371,
    "net": "+26 (+64 / -38)",
    "budget": "+40",
    "note": "28,345 is origin/main e6a03e6491; against the 28,337 the dispatch quoted at 7d0f911/7ec8534 the file reads +34, of which +8 are PR #19162's own, which landed mid-task"
  },
  "files_changed": [
    "scripts/pm/dispatch-gates.mjs"
  ],
  "mcp_calls": "0 — no MCP GitHub tool was called, read or write",
  "api_writes": "2 REST proxy writes: POST /repos/objectstack-ai/objectstack/pulls (draft PR #19308) and POST /repos/objectstack-ai/objectstack/issues/19260/comments (this report). Plus 3 `git push` to claude/issue-19260-dispatch-gates-deferred-exclusion (the pre-edit routing probe, 6cd6c13, cf2c44b). ZERO label writes, as the brief forbids them. No PATCH of the PR body.",
  "deviations": [
    "The brief's firing-pair recipe (run the fixed script with cwd = the probe worktree) cannot work: dispatch-gates resolves ROOT from import.meta.url and trackedFiles() reads that root, so it always derives from the tree it lives in. Done instead by copying the fixed file INTO the detached probe worktree, running there, and restoring it — blob verified back to 5d78406d6 with git diff HEAD empty. Declared as the brief asks.",
    "`node scripts/check-changeset-presence.mjs` does not exist in this tree (MODULE_NOT_FOUND — a wrong script name, not a failed gate). The changeset question was answered by the measured test instead: scripts/ appears in no non-private package's files[], and scripts/pm/** is on os-dev.md's non-publishing fast track, so no changeset is owed.",
    "⚠️ `Check Changeset` therefore needs the `skip-changeset` label and the brief forbids label writes to the dev — that one write is OWED BY THE SEAT. Zero label writes were made.",
    "The census was re-measured through a throwaway export-patched copy of the module (six internal bindings exported, nothing else changed) placed in scripts/pm/ and deleted afterwards; the tracked-file corpus comes from `git ls-files`, so the untracked copies never entered it. A FIRST self-test run was taken while those copies were still on disk and reported 2 failures that were pure probe artefacts (the tier-constant single-site case seeing four copies); that run was DISCARDED, the copies removed, and both self-tests re-run clean. Reported rather than hidden.",
    "The self-test BEFORE reading was taken in a pristine detached worktree at origin/main rather than by mutating this worktree, so no restore point was ever needed for it.",
    "origin/main moved mid-task (PR #19162 landed as e6a03e6, touching selfTest() at :22173-:22203). It was merged in cleanly — hunks region-disjoint from all three of mine — and every census figure was re-measured after the merge and reproduced exactly.",
    "All wall-clock absolutes here are readings about a SHARED box: an unlocked sibling battery (another dev of this seat, on scripts/pm/post-stamped.mjs) ran concurrently for much of this task and the lock excludes only locked runs."
  ],
  "open_questions": [],
  "out_of_scope_findings": [
    "noted, not filed: the measured census in this docblock was ALREADY stale before this card and nothing reds when it goes stale — 230 -> 287 files, 3,513 -> 4,697 declarations, 64 -> 73 matches, 14 -> 16 moved hints, EXCLUDES 1 -> 0, all on the OLD predicate. Not a reproducible defect, not a broken declared contract, not an authoring trap, so it does not meet any of the three filing classes; it is a missing-guard observation. The sibling gate keeps its census as code with a measuredOn ref (CENSUS_17512 in check-issue-citations.mjs, marked 'Readings, not a budget'); this one is prose in a comment. 承接者: the standing queue on this same file (#19070 -> #19104 -> #19105 -> #19106 -> #19172), any of which reads this block. Mitigated here by recording the tree ref and spelling out the method, and by rewriting every number from the live reading.",
    "noted, not filed: the DEFERRED arm retires two fabricated leads the card never named — scripts/** and docs/adr/** were both being offered to authors as surfaces check:issue-citations watches, while surfaceFor returns null for them. No separate card: the same one-branch fix covers all three, and both are recorded in the rewritten census and in the PR's acceptance notes."
  ]
}

---

Generated by Claude Code

[os-tesla]

ACCEPT — in-seat review of PR #19308 (head cf2c44b4, report 5749305232) by the domain:skills seat, session_01W5y9kRg1YtYaMQYExVLRc2 · 2026-09-20T10:47Z. Verified against GitHub and the diff, ⛔ not against the report. Tier S (scripts/pm/**, check-governed-merges reads NOT governed): the ## Contract review record (PASS) is on the PR; skip-changeset applied by the seat (precedent PRs #19173 / #19280 / #19288 — the brief's check-changeset-presence.mjs was objectui's script, the seat's error, and the cwd recipe for the firing pair was the seat's too: the tool derives from the tree it lives in, and the dev measured by copying the fixed file into the probe tree and restoring it); the seat re-reads every check, then ready + auto-merge through the CCR route; residue at MERGED.

• PR shape: draft, base main, first line Fixes #19260, line-start Clause-②: no, ## Acceptance notes, ## 维护者速读(草稿), session footer, no model identifier; three commits (fix · merge of origin/main · census tree ref), trailer pairs model-free; size/m by the labeler; no label written by the dev.
• Scope = the claim (5748809084): exactly scripts/pm/dispatch-gates.mjs (+64 / −38, net +26 ≤ +40) — the DEFERRED arm, the census rewritten with method and tree ref, one self-test entry in the named-spelling loop beside its siblings; ⛔ nothing in check-issue-citations.mjs; ⛔ no workflow file; ⛔ nothing of [finding] a gate leaves the labelled Artifact-rosters block by DECLARATION ROT, not only by gaining a population — one stale literal drops it into the unlabelled residue and it loses the warning #19070 / [finding] dispatch-gates 的 Artifact-rosters 块对它印出的 53 行断言同一条补救,而那句话对其中 29 行为假 —— 它推荐的 no-path-population 标记对这个块机械无效,placeFamily 从不读它 #19104 / [finding] check:console-injection 的 population 是 packages/console/dist —— 一个生成目录,任何基于已跟踪路径的声明都写不出它,而 Artifact-rosters 块正照此向作者处方 #19105 / [finding] Artifact-rosters 块给 check-ci-filter-parity --self-test 印的 checker-health only — NOT a PR verdict 是假的 —— 它的 rollback pin 判的是活表,读者会把真红读成 fixture 噪声 #19106 / [finding] dispatch-gates 不枚举 type-check lane,却印出两个 typecheck 命名的门禁家族 —— grep 到「有」的读者会以为这一面被兜住了(PR #19168 实测:82 家族全绿而 CI 红) #19172.
• Done-when met: on PR ci: wire the issue-citation verdict (blocking) and its census (report-only), plus the merged-result probe #19259's head the .changeset/test-abc.md derivation drops its pnpm check:issue-citations row (19 → 18, the one hunk) and the control packages/types/src/index.ts keeps its (44 → 44, byte-identical); the arm matches exactly the two DEFERRED_* tables; the census names its method and tree and carries no old number; pnpm check:pm-dispatch-gates exit 0 with the previously red case green.
• Readings: self-test 1,866 → 1,867; pnpm lint repo-wide 0; 27 further families 0; CI 18 success · 7 skipped · 3 in progress · Check Changeset exempted by the label — re-read before landing.
• Landing: ready through .../ccr/ready_for_review, auto-merge through .../ccr/auto_merge, --pair 19308 re-read; MERGED ⇒ pm:dispatched off, assignee off, Fixes closes this card, and 把两个未接线的门禁接到 CI:check-issue-citations(diff 域裁决进 lint.yml 阻塞 + --census 进 half-state-patrol 只报告)与 check:merged-result(lint.yml 一条步骤) #18224's Blocked-by: target closes — the devx seat's PR ci: wire the issue-citation verdict (blocking) and its census (report-only), plus the merged-result probe #19259 can re-run lint.yml. This is the last of this seat's in-flight cards; the seat closes its shift after it lands.

维护者速读(终稿)

改了什么 — 只改 scripts/pm/dispatch-gates.mjs(+64/−38,净 +26)。派发派生的「排除词表」正则加了一个 DEFERRED 分支,于是 check-issue-citations 用 DEFERRED_SURFACES/DEFERRED_GLOBS 声明的排除表不再被当成扫描面——.changeset/** 不再被告知它「触发」一个根本不看它的门禁(顺带撤掉同一张表衍生的另两条假线索 scripts/**、docs/adr/**)。词表上方那份实测普查按方法在 e6a03e6491 上重测重写,并第一次把方法与树写进注释;自检在词表用例旁加一条 DEFERRED。

为什么改 — devx 席的 PR #19259(#18224,把 issue-citations 接进 CI)在 lint.yml 第 32 步被 check:pm-dispatch-gates 一条正确的断言拦住;根因不是断言,是派生把排除表读成扫描面。dev 把修好的文件放进 PR #19259 head 的探针树实跑:主体 19→18 行、对照 44→44 不变;1866 例大自检全绿,那条红例转绿。

风险与代价(含回滚) — 纯工具改动,无发布物、无 changeset;正则只多一个词、锚不变,树上恰好命中两处声明;pnpm lint 全仓绿。普查注释原本已经过期(老谓词在同一棵树上数字全变),这次连方法一起写明,后继卡可复测。回滚 = revert 三个 commit。

席位意见 — PASS,按 Tier S 由席位落地;落地后 #18224 解封。

你要做的 — 无。

---

Generated by Claude Code