Two artifact-roster gates report a green that is not PR clearance — the pnpm script names resolve to --self-test only

Filed by Claude Code from a measurement handed back by a dev agent during the patch round on #15968.

The defect

node scripts/pm/dispatch-gates.mjs --commands prints an "Artifact rosters" block. Two of its entries are reported GREEN while measuring nothing about the PR under review:

pnpm check:partof-closing-keyword
pnpm check:single-claim-paths

Both pnpm script names resolve to a --self-test invocation only. Their green is a statement about the checker's own fixtures, not about the pull request.

The invocation that actually judges a PR is the bare node form, and on a developer machine it does not run at all:

node scripts/check-partof-closing-keyword.mjs
  → exit 2, "NOT WIRED … judged nothing"

node scripts/check-single-claim-paths.mjs
  → exit 2, "NOT WIRED — PR_NUMBER is not set"

So a seat harvesting the roster block sees two greens that read as PR clearance and are not. The correct reading of those two rows is NOT MEASURED (exit 2), and the roster presents the opposite.

Why this matters more than a cosmetic mislabel

This is the failure class this repo has now hit twelve times: a reading that cannot fail is indistinguishable from one that passed. A --self-test cannot fail because of anything in the PR, so its green carries no information about the PR — yet it sits in the same harvested list, in the same shape, as gates whose green does.

The repair principle that keeps being re-derived: name the axis your control discriminates on, and check that it is the axis that can fail. A row in a harvested gate roster should either judge the diff or say plainly that it cannot.

Reproduction

node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands
# read the "Artifact rosters" block; note the two pnpm rows above
# then compare against the bare node invocations, capturing the exit code
# BEFORE any pipe:
node scripts/check-partof-closing-keyword.mjs > /tmp/a.log 2>&1; echo $?   # 2
node scripts/check-single-claim-paths.mjs     > /tmp/b.log 2>&1; echo $?   # 2

The discriminator is the echoed command in the roster: for these two rows it ends in --self-test, for the rows whose green is meaningful it does not.

Not asserted

Whether this is by design is not settled here. The roster block is documented as sitting outside the reconciliation total, and it is possible the two --self-test entries are there deliberately as checker-health rows rather than as PR verdicts. If so, the ask is narrower: the block should distinguish a checker-health row from a PR-judging row, so a harvest cannot read one as the other.

No grading of domain:* or type is applied here — that is triage's call.

分诊 · `domain:devx` / `bug` / `tooling` / `priority:p2` / `pm:queue`

分诊席位。⛔ 不认领、不派发、不写代码、不合并。origin/main @ 932acc3d，2026-09-06T04:31Z。

复现 —— 两个 pnpm 脚本名字面上以 `--self-test` 结尾

package.json
  "check:partof-closing-keyword": "node scripts/check-partof-closing-keyword.mjs --self-test",
  "check:single-claim-paths":     "node scripts/check-single-claim-paths.mjs --self-test",

两个裸 node 脚本也都存在（scripts/check-partof-closing-keyword.mjs、scripts/check-single-claim-paths.mjs）。

⇒ ⭐ 卡给的鉴别器是可机检的：roster 里回显的命令以 --self-test 结尾的行，其绿色评的是检查器自己的夹具；不以它结尾的行才评 PR。⛔ 这不需要读脚本源码就能判定，也就是说修法可以是机械的。

⚠️ 我没有跑那两条裸 node 调用（卡报的 exit 2 / "NOT WIRED … judged nothing" / "PR_NUMBER is not set"）——本容器没有 PR_NUMBER 环境，跑了也只会复现「未接线」。按卡的标注采信。

定型 `bug` / 定级 p2

bug：roster 把一条结构上不可能因 PR 而失败的绿色，与评 PR 的绿色并排、同形地呈现。⇒ 呈现与含义不符，恢复方向。

p2，理由是卡自己给的类名与计数：

This is the failure class this repo has now hit twelve times: a reading that cannot fail is indistinguishable from one that passed.

⇒ 而且它作用在收割点上——席位是按这个 roster 块批量读取门禁结论的。⭐ 一个假绿单独存在只误导一个人；一个假绿坐在被批量收割的清单里，会误导每一个收割它的人。

不给 p1：没有错误代码因此合入（就已知而言），也没有产品缺陷。

车道 `domain:devx`

落点 scripts/pm/dispatch-gates.mjs（roster 块的呈现）与/或 package.json 的两个脚本名 ⇒ 车道表 scripts/ gates 在 domain:devx 行。

⭐ 修法：两条路，都可执行，不需要裁决

卡把「是否 by design」留开了，并给了对应的窄 ask。我判定两条路的交付物是同一个，所以 pm:queue 而非 needs-user-decision：

若这两行是刻意的 checker-health 行 ⇒ roster 必须把 checker-health 行与 PR-judging 行区分开，让收割不可能把一个读成另一个；
若不是刻意的 ⇒ 把它们改成评 PR 的裸 node 形式，届时它们会诚实地报 exit 2 / NOT MEASURED。

⇒ 无论哪种，都要让这两行不再冒充 PR 清白。⛔ 接卡人可以自己判定是哪种（读 roster 块的 docblock 即可），⛔ 不必上交裁决。

验收：⭐ 卡给的修复原则就是验收标准——

name the axis your control discriminates on, and check that it is the axis that can fail. A row in a harvested gate roster should either judge the diff or say plainly that it cannot.

相邻 —— ⛔ 不合并，但强烈建议同一个人一次做完

#16055（本轮我已路由 domain:devx / enhancement / p2）—— 「没有任何闸门判 changeset 的 LEVEL 是否配得上它移动的面；同一个 diff 在 patch 和 minor 上 Check Changeset 全绿」。⭐ 同一个类的第 13 例，而且 #16055 自己就点名了本卡：

This is a repeat of the shape already recorded in #16030（two roster gates whose green grades the checker's own fixtures rather than the PR）。

⇒ 两张卡是同一条原则的两个实例，⛔ 不合并（不同文件、不同闸门），但同批做时那条原则只需要想一次。
⚠️ 而且值得考虑把那条原则写进闸门写作规范（skills 车道）——本轮我在 #16055 上也提了同一句。⛔ 那是第三张卡的事，不在这两张的范围内。

Generated by Claude Code

Claim: PM loop round 1
Session: session_01Vbw3RPgdtqesx4azk9SbW8
Branch: claude/issue-16030-roster-self-test-rows
Worktree: objectstack-issue-16030
Domain: domain:devx
File surface: scripts/pm/dispatch-gates.mjs (the Artifact rosters block and, for rider #12797, the extensionless-target reason ordering) and scripts/check-cross-package-test-inputs.mjs (rider #12808, one docblock clause). Root package.json is NOT on this surface: it is held by PR #15392 (another account) — if the fix needs the two script names changed there, stop and report. (stop on breach; explain in the report)
Container & model: M (one card plus two pre-declared hold riders), mode:subagent, model: opus — node scripts/pm/dispatch-gates.mjs --tier at 33388f9 (05:42Z): no path-derived mandate, floor sonnet · default opus · ceiling fable
Clause-②: no
Serial constraints cleared: open-PR file map taken 2026-09-06T05:41Z: no open PR touches scripts/pm/dispatch-gates.mjs or scripts/check-cross-package-test-inputs.mjs. H17 riders that fire on this surface, both folded into this dispatch per their own Restart-when: lines: #12797 (hold 5445380987, Restart-touch: scripts/pm/dispatch-gates.mjs) and #12808 (hold 5445950277, Restart-touch on both files). Each rider card carries its own pointer claim to this branch; the PR closes all three with one Fixes line per card.

Generated by Claude Code

os-dev-report

{
  "issue": 16030,
  "status": "done",
  "branch": "claude/issue-16030-roster-self-test-rows",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/16221",
  "premise_still_valid": true,
  "summary": "Anchor #16030 plus both pre-declared riders #12797 and #12808, all three in PR #16221. #16030 took route (i), the checker-health split, and the route is settled by evidence rather than taste: the two named rows are self-test-only ON PURPOSE, because the half of each gate that judges a pull request needs the pull_request event payload and runs in its own workflow (partof-closing-keyword-guard.yml, single-claim-path-guard.yml) while lint.yml runs the checker-health half. Root package.json therefore needed no change and was not touched, so the declared blocked case did not fire. The Artifact rosters block now splits its rows, derived not hand-kept, with the marker ON THE ROW (a harvest greps rows, per spellingDistribution's own measurement) and a third answer, null, for an invocation this tool cannot resolve rather than defaulting it into either side. Split measured on this tree: 7 checker-health, 32 PR-judging, 0 unresolved; the same axis is carried in --commands stderr and in artifactRosterSilences in --json so the three readings cannot disagree. #12797: the bare-word refusal is now ordered ahead of the extensionless-target branch so the printed reason is the refusal that fired, and that refusal has ONE owner (refusedAsTooGeneric) which hintCovers itself calls; the target sentence was NOT taught which refusal fired, per the ruling. #12808: one docblock clause in check-cross-package-test-inputs.mjs completes the single-segment refusal's statement of the rule with its one exception (moduleRelativeDirectoryHint) and why an import specifier is not in it. No code change.",
  "tests": "All 33 derived families green at e6e354bc; --ran reconciles 33 derived / 33 run / 0 NOT-MEASURED / 0 UNRUN. Every exit code captured before any pipe (cmd > log 2>&1; EXIT=$?). See `gates`. Self-test cases added: 16 for #16030 (the predicate on both carriers and on the conventional --self-test && work shape; the rendering split; the row-level marker; the unclassified bucket) and 7 for #12797. ABLATION, four legs, each proving the mutation reached disk before any reading was taken (anchor count 1 -> 0, injected text present, `git hash-object` blob moved off the HEAD blob) and each restore proven by blob equality to the HEAD blob AND an empty `git diff HEAD`, restoring with `git checkout HEAD -- PATH` (never a bare `git checkout --`), under a `trap ... EXIT INT TERM` with absolute paths. dispatch-gates.mjs is run from source, not from a dist, so there is no build leg to interpose. Leg 1 every() -> some() in selfTestOnlyInvocation: the conventional-two-segment case reds (1 fail / 6 pass). Leg 2 classify pnpm rows from the PRINTED command, i.e. the card's own stated discriminator: the manifest-body case AND the null case red (2 fail / 5 pass) - this is the falsification, executed rather than argued. Leg 3 marker moved off the row, caption kept: the row-marker case reds (1 fail / 6 pass). Leg 4 #12797 ordering reverted: both bare-hint reason cases red (2 fail / 5 pass). Un-ablated control before all four: 7 pass / 0 fail. REACHABILITY re-measured for #12797 with the hold comment's own method, before and after: 236 dead hints on this tree, 0 reaching the `target` branch, 0 of those bare - still LATENT, same reading as 2026-08-27 at ead731756. LINT narrowed to the two touched files and the narrowing proven rather than asserted: receiving population read from eslint.config.mjs itself via ESLint#isPathIgnored over the tracked corpus = 6189 of 7951 tracked files; files linted read from --format json = 2, 0 errors, 0 warnings, exit 0 at e6e354bc; invariance = eslint.config.mjs states in its own prose, with a measured positive control, that this repo never enables type-aware linting for ANY file, so nothing in this diff can move the verdict on an untouched file. Control-byte scan on every GitHub-bound text: `grep -naP '[\\x00-\\x08\\x0b\\x0c\\x0e-\\x1f\\x7f]'` found none; check:nul-bytes green.",
  "mcp_calls": "1 - one targeted mcp search_issues for the out-of-scope dedup. /search/* is refused by the egress proxy by design (rest-channel.md), so search is the one read with no REST equivalent; every other read and write on this run went through repo-scoped REST.",
  "gates": [
    "node packages/lint/scripts/check-reference-carrier-shape.mjs -> exit 0 @ e6e354bc | check-reference-carrier-shape: OK — 6193 file(s) scanned, 602 `reference` site(s).",
    "node packages/lint/scripts/check-reference-carrier-shape.mjs --self-test -> exit 0 @ e6e354bc | ✅  self-test: flags an object / array / numeric carrier at a field-def position in all",
    "node scripts/check-ci-filter-parity.mjs -> exit 0 @ e6e354bc | OK: all 168 declared cross-package glob(s) (119 unique) are covered by `core` or `crosspkg`, every `crosspkg` entry still covers one, and the `test` j",
    "node scripts/check-closing-keyword-parity.mjs -> exit 0 @ e6e354bc | check-closing-keyword-parity: OK (3 parsers agree on all 9 keywords and both measured separators; sweep found 5 file(s) carrying the grammar across 79",
    "node scripts/check-closing-keyword-parity.mjs --self-test -> exit 0 @ e6e354bc | ✓ check-closing-keyword-parity --self-test: 24 assertions, 5 mutations of the shipped parsers each driven to red.",
    "node scripts/check-comment-mask-corpus.mjs -> exit 0 @ e6e354bc | ✓ comment-mask corpus sweep [scripts/js-comment-mask.mjs]: 6193 files, 0 disagree, 0 unparseable, 105.1s (comparator self-test: 17 cases pass).",
    "node scripts/check-declaration-mirrors.mjs -> exit 0 @ e6e354bc | OK: 6 hand-written declaration(s) agree with their modules on name, kind and required arity.",
    "node scripts/check-declaration-mirrors.mjs --self-test -> exit 0 @ e6e354bc | All 29 self-test cases passed.",
    "node scripts/check-published-list-mirrors.mjs -> exit 0 @ e6e354bc | OK: 1 published list mirror(s) match their constants line for line.",
    "node scripts/check-published-list-mirrors.mjs --self-test -> exit 0 @ e6e354bc | All 37 self-test cases passed.",
    "node scripts/check-scripts-symbol-anchors.mjs -> exit 0 @ e6e354bc | ✅ check-scripts-symbol-anchors: 2786 anchors across 220 scripts resolve — 21 symbol (21 declaration, 0 literal), 2765 file-level, 0 cross-repo, 1 ",
    "node scripts/check-scripts-symbol-anchors.mjs --self-test -> exit 0 @ e6e354bc | ✅ check-scripts-symbol-anchors --self-test: every finding class provoked, comment-prose projection wired, declined shapes counted not missed, allowa",
    "node scripts/check-self-test-wired.mjs -> exit 0 @ e6e354bc | ✓ check-self-test-wired: every one of the 179 script(s) CI runs that ship a `--self-test` has that self-test run by CI.",
    "node scripts/check-self-test-wired.mjs --self-test -> exit 0 @ e6e354bc | check-self-test-wired --self-test: 4 live ledger row(s) verified, plus the comment mask, the right boundary, alias resolution and both audit direction",
    "node scripts/check-self-test-workflow-commands.mjs -> exit 0 @ e6e354bc | ✓ check-self-test-workflow-commands: no self-test CI runs prints a line the Actions runner would parse as a workflow command.",
    "node scripts/check-self-test-workflow-commands.mjs --self-test -> exit 0 @ e6e354bc | check-self-test-workflow-commands --self-test: both measured parse rules pinned (legacy form anywhere in a line, current form only at line start), the",
    "node scripts/check-whole-set-label-write.mjs -> exit 0 @ e6e354bc | ✓ check-whole-set-label-write: 0 violations — 274 file(s) over 3 root(s) · 11 raw mention(s) · 11 in comments/prose (cleared) · 0 in EXECUTABLE",
    "node scripts/check-whole-set-label-write.mjs --self-test -> exit 0 @ e6e354bc | ✓ check-whole-set-label-write --self-test: all cases pass (24 fixture trees + 5 refusals + 1 allowlist hatch)",
    "node scripts/pm/bare-root-worklist.mjs --self-test -> exit 0 @ e6e354bc | OK  self-test: 78 live row(s), 57 unreachable as spelled, 46 recorded verdict(s) — none stale, none missing, none contradicted (12 row(s) whose gate",
    "pnpm check:agent-test-spelling -> exit 0 @ e6e354bc | ✓ check-agent-test-spelling: 0 violations — 459 file(s) · 6913 bare `--` token(s) · 1533 launcher-rooted run(s) · 9 separator(s) JUDGED · 5 vi",
    "pnpm check:bash32-floor -> exit 0 @ e6e354bc | ✓ check-bash32-floor: 27 tracked shell file(s) under scripts/**, .claude/hooks/**, .githooks/** name no bash 4+ construct outside a comment, a guard",
    "pnpm check:cli-command-ids -> exit 0 @ e6e354bc | ✓ check-cli-command-ids: 427 command-id literal(s) across 125 file(s) outside packages/cli all resolve to a real command path (73 ids derived; 3 dec",
    "pnpm check:cross-package-test-inputs -> exit 0 @ e6e354bc | OK: 27 package(s) read outside themselves, all declared, and turbo.json hashes every declared glob.",
    "pnpm check:declared-population-live -> exit 0 @ e6e354bc | ✓ check:declared-population-live — 223 of 268 famil(ies) declare a path population, and every one of them reaches this tree's 7951 tracked file(s)",
    "pnpm check:driver-memory-census -> exit 0 @ e6e354bc | check-driver-memory-census: OK — every declaration is ledgered, every ledger entry is live, and every ruled file states \"#6664 census: 2 ruled consu",
    "pnpm check:entry-guard -> exit 0 @ e6e354bc | ✓ check:entry-guard: 220 scripts/ file(s) — every entry guard goes through invoked-as.mjs; 164 export bindings, 164 of them inert on import (0 kno",
    "pnpm check:nul-bytes -> exit 0 @ e6e354bc | check-nul-bytes: OK (scanned 7944 text file(s) -- 7944 tracked, 0 untracked-not-ignored; skipped 7 binary; no raw ASCII control bytes).",
    "pnpm check:parse-guard -> exit 0 @ e6e354bc | ✓ check:parse-guard: 219 scripts/ file(s) — every TypeScript parse goes through ts-parse.mjs.",
    "pnpm check:pnpm-filter-targets -> exit 0 @ e6e354bc | ✓ check:pnpm-filter-targets: 144/186 `--filter` occurrence(s) across 35 file(s) resolve against 79 workspace package(s); 42 not judged (2 foreign, 2",
    "pnpm check:ratchet-remedy-authority -> exit 0 @ e6e354bc | OK  check-ratchet-remedy-authority: 212 scripts swept (scripts/*.{mjs,mts} + scripts/pm/*.{mjs,mts}); 14 mark the expanding remedy ⛔ MAINTAINER-ONLY",
    "pnpm check:refd-timer-probe -> exit 0 @ e6e354bc | OK  check-refd-timer-probe: 6188 source file(s) swept; the process-global timer probe is read in packages/qa/refd-timer-testkit/src/index.ts and nowhe",
    "pnpm check:watch-hint-literal -> exit 0 @ e6e354bc | ✓ check-watch-hint-literal: 62 declaration(s) across 4 rostered name(s) -- ROOT_DIR_WATCH_HINTS 40, ROOT_FILE_WATCH_HINTS 11, ROOT_WATCH_HINTS 3, DE",
    "pnpm check:pm-dispatch-gates -> exit 0 @ e6e354bc | \"✓ dispatch-gates self-test: 1534 cases pass.\" (under scripts/pm/os-verify-lock.sh, OS_VERIFY_LOCK_SLOT=issue-16030; \"os-verify-lock: VERDICT command-exit 0 · held the lock 773s (12m53s) · waited 517s (8m37s)\"). Detached per this file's own header (its battery exceeds the container foreground cap) and blocked on in-round. First run at the WIP commit was 1533/1534 with the one failure being this diff's owed COMPOUND_ANCHOR_LEDGER row; that row was added and this is the re-run.",
    "node scripts/pm/dispatch-gates.mjs --changed --ran RANFILE (all 33) -> exit 0 @ e6e354bc | \"Run reconciliation — 33 derived, 33 run, 0 NOT-MEASURED, 0 UNRUN.\"",
    "pnpm exec eslint --no-inline-config --format json (the 2 touched files) -> exit 0 @ e6e354bc | 2 files, 0 errors, 0 warnings"
  ],
  "files_changed": [
    "scripts/check-cross-package-test-inputs.mjs",
    "scripts/pm/dispatch-gates.mjs"
  ],
  "line_budget": "n/a",
  "deviations": [
    "NOT A DEVIATION but a falsification worth the PM's attention: the card's stated discriminator ('the echoed command ends in --self-test') is WRONG for the two rows the card was filed for. Measured over the 39 roster rows on this tree it selects five rows and misses both named ones, because they are pnpm-spelled and carry the flag in the manifest entry that defines the name, never in the printed bytes. A fix implementing the card's literal discriminator would have shipped a control that cannot fail for the population it exists for - the same defect one level up. Ablation leg 2 executes that wrong implementation and shows it reds the manifest-body case. PM assumption A1 is therefore half right: package.json does carry --self-test for both, but the roster's rendered text does not.",
    "SCOPING, declared: residueLines' count summary was NOT extended with a checker-health count. It has strict derivability refusals and ~15 self-test call sites, which is disproportionate for a count the roster block itself already prints; the per-row axis is in --json instead.",
    "A3 partly falsified in the cheap direction: pnpm check:pm-dispatch-gates DID complete here, twice, under the shared lock - 650s and 773s held, 524s and 517s waited. No battery had to be narrowed and nothing is NOT MEASURED. It was run detached per the file's own header (its battery exceeds the container foreground cap) and blocked on in-round with a bounded foreground wait, never left to a background watcher.",
    "The first full self-test run reported one real failure, now fixed in the same PR: the new export selfTestOnlyInvocation trips the compound self-test name anchor and owes a COMPOUND_ANCHOR_LEDGER row. Classified accidental: true and added, which is what that table's 'Maintaining this table' section requires; it explicitly forbids renaming the function to dodge the anchor.",
    "No off-surface touch. Root package.json was NOT edited (correctly - the two script definitions are right as they are), so the blocked case declared in the dispatch did not fire. Files changed are exactly the two on the declared surface.",
    "origin/main moved during the run (0c5d0359 -> 159dbad2). dispatch-gates printed its STALE TREE advisory naming four files it derives from; none of them is a family for either of my paths, and the derived list is unchanged. The branch is based on 0c5d0359 and the change set is three-dot against the merge base.",
    "PR body carries two attribution footers: the one the harness attribution instruction requires, and the session-URL block the platform itself appended on creation. Both name the same session. See open_questions."
  ],
  "open_questions": [
    {
      "question": "#12797's ruling leaves retiring the `target` branch as an optional follow-on and says to report rather than assume. Reported: on this tree that branch is reached by ZERO dead hints (236 swept), and after this reorder its only remaining reachable population is separator-carrying hints, which hintCovers structurally never leaves dead since #12514. So it now has no live caller in either population.",
      "options": [
        "A - keep it, as shipped: it stays pinned by a pure-function self-test case, and the cost is one dead branch",
        "B - retire it in a follow-on card, which also touches unreachableClass (which reads `target`) and the extensionlessModuleTarget pins"
      ],
      "recommendation": "A here, B as a separate card. The rider was ruled to be a small edit and retirement is a wider blast radius than one reason branch; the reading is now recorded with a number so the follow-on does not have to re-derive it."
    },
    {
      "question": "Attribution footer conflict, flagged rather than silently resolved. The dispatch contract says the PR body ends with the session-URL footer form; the harness attribution instruction states a different spelling and says it replaces earlier attribution guidance. I wrote the harness form, and the platform then appended its own session-URL block, so the body carries both.",
      "options": [
        "A - leave both (as shipped): the required attribution is present and the platform's canonical block is intact",
        "B - PATCH the body to drop the hand-written block, risking the platform replacing the session-URL form with the bare form on edit"
      ],
      "recommendation": "A. Both blocks name the same session, nothing is misleading, and editing to tidy it risks losing the session URL - which is the half that actually carries provenance."
    }
  ],
  "out_of_scope_findings": [
    "NOT filed, already open as #15310: COMPOUND_ANCHOR_LEDGER's docblock census ('all 22 / fifteen genuine / seven accidental') no longer counts its own table, which held 28 rows before this PR and 29 after. Found while adding this diff's owed ledger row. Deduped with one targeted search including closed cards; #15310 is open, labelled finding / priority:p3 / domain:devx, and names exactly this drift. Not repaired in-place: re-taking that census is not a mechanical edit, so the bounded in-place exemption does not apply."
  ]
}

Generated by Claude Code

ACCEPT — anchor #16030 and riders #12797 / #12808, all on PR #16221. PM seat domain:devx @ objectstack, session session_01Vbw3RPgdtqesx4azk9SbW8, 2026-09-06T07:0xZ. Reviewed against GitHub (PR #16221, its diff and files, check-runs on head e6e354b), not against the report's narrative. Report: 5557586086 (the 06:55Z os-dev-report on this card); rider pointers posted 06:56Z on #12797 and #12808.

Checklist conclusions

PR shape: draft, base main, first three lines Fixes #16030 / Fixes #12797 / Fixes #12808, no other closing keyword; footer present (twice — the harness form and the platform-appended session-URL block; both name this session, accepted as-is, open question 2 = A).
Scope (get_files): exactly scripts/pm/dispatch-gates.mjs (+353/−16) and scripts/check-cross-package-test-inputs.mjs (+7/−2); root package.json untouched (the declared blocked case did not fire — correctly: the two check:* script definitions are right as they are). skip-changeset on the PR. Governed test on that file list: NOT governed ⇒ ordinary queue.
Diff spot-read by this seat: (Two artifact-roster gates report a green that is not PR clearance — the pnpm script names resolve to --self-test only #16030) the roster classifies each row from the invocation the manifest entry DEFINES (selfTestOnlyInvocation over entry.manifestCommand, or the direct command), three outcomes — PR-judging / checker-health / null unresolved — with the marker on the row itself and the same axis carried into --json as artifactRosterSilences[].checkerHealth; ([finding] dispatch-gates residue: after #12514 the extensionless-target reason is only reachable for a BARE hint, where it names the wrong cause — latent wrong-reason row, zero instances today #12797) refusedAsTooGeneric is the single owner of the bare-word refusal and is consulted ahead of the extensionless-target sentence, the target sentence is not taught which refusal fired, and a separator-carrying hint still gets the extensionless sentence byte for byte (pinned); ([finding] check-cross-package-test-inputs' docblock states the single-segment hint refusal without its one exception #12808) one docblock clause names the moduleRelativeDirectoryHint exception and why an import specifier scores as no population structurally. No code change on [finding] check-cross-package-test-inputs' docblock states the single-segment hint refusal without its one exception #12808.
Evidence: 33 derived families run at e6e354b, 33/33 reconciled with --ran, all exit 0 including pnpm check:pm-dispatch-gates (1534 cases; lock VERDICT command-exit 0 · held 773s · waited 517s); four ablation legs each proved on disk (blob moved off HEAD) and restored by blob equality + empty git diff HEAD; [finding] dispatch-gates residue: after #12514 the extensionless-target reason is only reachable for a BARE hint, where it names the wrong cause — latent wrong-reason row, zero instances today #12797 reachability re-measured with the hold comment's method (236 dead hints, 0 reach target, still latent). CI on head: 34 checks complete, 23 success / 11 skipped / 0 failing.

⭐ The card's own discriminator was falsified, and the dev did the right thing. The card said the tell is "the echoed command ends in --self-test". Measured over the 39 roster rows, that literal selects five rows and MISSES both rows the card was filed for, because they are pnpm-spelled and the flag lives in the manifest entry, never in the printed bytes. Ablation leg 2 executes the card's literal discriminator and shows it reds the manifest-body case. This seat's Zone-2 A1 was half right (the manifest carries the flag) and half wrong (the rendered text does not); the PM's assumption is corrected here in public, and the card's premise is corrected by this comment rather than by editing the body.

Open questions: Q1 (retire the target branch) = A, keep it pinned as shipped — the rider was ruled a small edit and retirement widens into unreachableClass and the extensionless pins; the zero-reachability number is now on record on #12797 so a future card need not re-derive it; ⛔ no follow-on card is filed by this seat (optional follow-on, no live caller, no user impact). Q2 (double footer) = A.
Out-of-scope: #15310 (COMPOUND_ANCHOR_LEDGER census drift, 28→29 rows) — already open, correctly not repaired in place.

Landing: non-governed, all checks complete on the ACCEPT head ⇒ this seat flips ready, arms auto-merge (squash) and confirms added_to_merge_queue as the next act. On MERGED, Fixes closes all three cards; this seat strips pm:dispatched and clears the assignee on each in the landing stroke.

Generated by Claude Code

LANDED — PR #16221 MERGED via the merge queue as a3bbb8b1a at 2026-09-06T07:25:38Z (added_to_merge_queue 07:01:32Z → removed_from_merge_queue + merged one second apart = a landing). PM seat domain:devx @ objectstack, session session_01Vbw3RPgdtqesx4azk9SbW8, 2026-09-06T07:39Z.

Content probe on re-fetched origin/main @ 618f70d74 via git show: export function selfTestOnlyInvocation = 1, the row marker checker-health only (--self-test) — NOT a PR verdict = 1, artifactRosterSilences (the --json axis) = 2 — all in place in scripts/pm/dispatch-gates.mjs; firing control Artifact rosters — = 3; nonsense token = 0.

Same stroke: pm:dispatched stripped and the assignee cleared (GitHub closes the card through Fixes and leaves labels — #16005). Ownership labels stay.

Generated by Claude Code

Navigation Menu

Repository navigation

Two artifact-roster gates report a green that is not PR clearance — the pnpm script names resolve to --self-test only #16030

The defect

Why this matters more than a cosmetic mislabel

Reproduction

Not asserted

分诊 · `domain:devx` / `bug` / `tooling` / `priority:p2` / `pm:queue`

复现 —— 两个 pnpm 脚本名字面上以 `--self-test` 结尾

定型 `bug` / 定级 p2

车道 `domain:devx`

⭐ 修法：两条路，都可执行，不需要裁决

相邻 —— ⛔ 不合并，但强烈建议同一个人一次做完

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Two artifact-roster gates report a green that is not PR clearance — the pnpm script names resolve to --self-test only #16030

Description

The defect

Why this matters more than a cosmetic mislabel

Reproduction

Not asserted

Activity

os-zhuang commented on Sep 6, 2026

分诊 · domain:devx / bug / tooling / priority:p2 / pm:queue

复现 —— 两个 pnpm 脚本名字面上以 --self-test 结尾

定型 bug / 定级 p2

车道 domain:devx

⭐ 修法：两条路，都可执行，不需要裁决

相邻 —— ⛔ 不合并，但强烈建议同一个人一次做完

claude commented on Sep 6, 2026

claude commented on Sep 6, 2026

claude commented on Sep 6, 2026

claude commented on Sep 6, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

分诊 · `domain:devx` / `bug` / `tooling` / `priority:p2` / `pm:queue`

复现 —— 两个 pnpm 脚本名字面上以 `--self-test` 结尾

定型 `bug` / 定级 p2

车道 `domain:devx`