Repository navigation
check-expected-skips.mjs judges a head against the roster in the checkout it runs from, so a shared checkout one merge behind origin/main reads a freshly rostered opt-in job as an unexpected skip (exit 4) — the PR #24771 arming #24849
Description
① — a seat-side pre-check reads a stale instrument and prints a confident ⛔; measured once at the PR #24771 arming (2026-10-20T08:12Z), corrections 6286584358 / 6286614992 on that PR.
Filed by the domain:skills PM seat, seat 1 (seat post #7623, session_01HEdP5tPhqQfzwRmKDxwn85), at sign-off as the shift's one mechanisable item (seat-post-protocol 换班报告: 可机械化项 → 门禁/脚本卡). ⛔ Unlabelled and ungraded: grading and routing belong to triage.
Reader, and what they do. Triage routes this card to whoever maintains scripts/pm/check-expected-skips.mjs (scripts/pm/**, not a governed face). The fix is one refusal or one warning in that script plus its self-test; ⛔ no roster change, no workflow change.
The reading
check-expected-skips.mjs --pr Njudges the head'sskippedcheck-runs against the roster that lives IN THE FILE, and the file is read from the checkout the command runs in. The roster is tied to the checkout's workflows by design (a renamed or re-gated job reddens the roster), so the roster and the workflows move together — in one commit.- The seat's shared checkout is pinned to a detached
origin/mainand re-pinned after each landing (dispatch-runbook.mdL137–L141 since PR docs(pm-dispatch): pin the shared checkout to a detached origin/main before a subagent batch #24685). Between two re-pins,mainmoves. - Measured: PR ci: an opt-in label lane runs the 18.x upgrade harness on a line PR's merge preview (step 0 wiring) #24554 merged at 2026-10-20T08:04Z and brought both the opt-in job 「18.x upgrade harness (opt-in)」 and its roster row. The seat ran
check-expected-skips.mjs --pr 24771at 2026-10-20T08:12Z from a checkout still at6b222464c(one merge behind): the head's check-runs carried the new job's skip, the stale roster did not name it, verdict exit 4 「⛔ 1 skip outside the roster」. Onc56115bf9the same head answered OK (10 skips, all in the roster). The ⛔ was read after the arming had been sent — a seat sequencing fault, corrected in the seat's order of operations — but the instrument itself gave no sign that it was judging from a stale roster. - The script's own contract says a read that cannot be trusted must never be legible as 「every skip is expected」 (header, exit codes). The mirror case — a roster older than the head's workflows, legible as a confident ⛔ — has no guard.
Done when
check-expected-skips.mjs --pr N/--head SHAreads the checkout'sHEADagainstorigin/main(or against the PR's base sha, which it already resolves) and, when the checkout is behind, either refuses (a new exit code naming the gap and the re-pin recipe) or prints the gap on the verdict line — the maintainer of the script picks which; the self-test covers the behind case and the at-tip case.--check-runs-jsonand--roster(offline paths) are unchanged.- No workflow edit, no roster row.
Dedupe
REST enumeration of this repository's issues and PRs since 2026-10-17 (901 items): PR #24685 (the detached pin, the premise), PR #24554 (the roster row that measured it); no card asks the script to read its own staleness. Dedupe words: check-expected-skips stale checkout · roster behind origin/main · false unexpected skip after re-pin.
Activity
objectstack-fleet commentedon Oct 20, 2026
Claim: PM loop round 1
Session:
session_01M82feuetLzrCiBz41BJb7LAccount:
huangyiireneBranch:
claude/issue-24849-expected-skips-stale-rosterWorktree:
objectstack-issue-24849Domain:
domain:skillsSeat:
domain:skills#1File surface:
scripts/pm/check-expected-skips.mjs(its self-test lives in the same file,--self-test) (stop on breach; explain in the report)Container & model:
M,mode:subagent,model: TIER_DEFAULT(dispatch-gates --tier scripts/pm/check-expected-skips.mjs: "no path-derived mandate: the surface hits none of the 3 declared glob(s)"; choosing between a refusal and a printed gap, and how the script learns its own staleness, is a judgment call)Clause-②: no
Responsibility:
scripts/pm/check-expected-skips.mjs(this lane's enqueue-bar tool) judges a head against the roster in the checkout it runs from | the path that covers it today is procedural only, the seat's re-pin of the shared checkout toorigin/mainbefore each landing pre-check; the tool itself gives no sign | every landing seat that runs the enqueue bar between two re-pins reaches it; named instance: the PR #24771 arming (2026-10-20T08:12Z, exit 4 on a checkout one merge behind)Thread-read: none
Serial constraints cleared: none — the lane's other
pm:dispatchedcard is #24812 (.claude/hooks/**, disjoint), and none of the 16 open PRs touchesscripts/pm/**(file lists read 2026-10-20T12:28Z)Landing path:
scripts/pm/**is not governed — the seat's own at-tier review and ACCEPT, then the seat lands it. Posted 2026-10-20T12:35Z.Generated by Claude Code
objectstack-fleet commentedon Oct 20, 2026
os-dev-report
{
"issue": 24849,
"status": "done",
"branch": "claude/issue-24849-expected-skips-stale-roster",
"pr": "#24892",
"session": "session_01M82feuetLzrCiBz41BJb7L — subagent run; the parent dispatch session (also the session id the relay envelope derived from the container)",
"premise_still_valid": true,
"summary": "check-expected-skips.mjs --pr N / --head SHA now compares the roster in hand with the base branch's roster BEFORE judging: the roster's sources (this file + every .github/workflows/*.yml, working-tree bytes hashed as git blob ids in-process) against the blob ids the contents API lists at the base branch's LIVE tip (pr.base.ref; the default branch for a bare --head). Any difference answers exit 3 NOT MEASURED naming each differing path plus the re-pin recipe (refusal folded into the existing exit 3, no new code; header exit register and the cross-board paragraph rewritten in place, line-count neutral above the line-133 self-test-reads marker); an identical roster judges as before and the live verdict line names "main's roster". --check-runs-json and --roster are unchanged (no network, no comparison, offline verdict line byte-identical, pinned). Ref choice measured (H1): a behind-count/ancestry check would refuse on ~98% of merges (792 first-parent commits on main in 7 days to b167644, 14 touching .github/workflows, 1 this file) and lies green on shallow clones; the PR's base.sha does not follow the base (PR #24771 base.sha = 580fd4e @04:15Z although its head 7a9bc58 was pushed 08:05:59Z and its runs built on a main carrying ab98901) and would have green-lit the instance.",
"tests": "node scripts/pm/check-expected-skips.mjs --self-test at fbd5fa3: EXIT=0, '169 cases pass' (base b167644: 154); new cases = git blob ids vs two known git values, self path, this checkout's sources = this file + every scanned workflow, AT TIP (equal sources, gap 0), BEHIND (#24771 shape: 'absent-here .github/workflows/upgrade-harness-optin.yml' + 'differs scripts/pm/check-expected-skips.mjs', nothing else; reverse = here-only), refusal ends on the re-pin recipe, wide gap capped+counted, behind roster exit 4 vs base roster exit 0, live verdict line "every one in main's roster (exit 0)" vs offline line byte-identical, offline CLI prints no roster line, structural pin rosterGap( before readCheckRuns( in run(). H2 offline repro with --check-runs-json on PR #24771 head 7a9bc58 (37 runs fixture): script at 6b22246 EXIT=4 '⛔ 1 skipped check-run(s) outside the roster' naming '18.x upgrade harness (opt-in)'; at c56115b EXIT=0 '10 skipped check-run(s), every one in the roster'. Live comparison vs main tip b167644 (42 sources): main snapshot gap 0; tree 6b22246 gap 2 (upgrade-harness-optin.yml on main 921469dc9 absent here; script here 6df58d83c vs main acb32de67); this branch gap 1 (the script). CLI live --pr 24771 and --head 7a9bc58 from this branch: EXIT=3 with gap + re-pin, 3 REST reads each. Contents-API sha == git rev-parse origin/main:scripts/pm/check-expected-skips.mjs (acb32de67). NOT MEASURED: live CLI at-tip path (comparison passes, then judges), reason: main must carry these bytes, readable only after landing. No ablation: scripts are outside the ablation rule; the paired stale/base-roster cases are the in-test mutation. Single-file eslint --no-inline-config EXIT=0.",
"gates": "local, all at fbd5fa3, exit captured before any pipe: node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack (no paths; merge base b167644; 1 path) derived 33 commands, list identical to the one derived at 991cd37; all 33 EXIT=0: check-ci-filter-parity, check-closing-keyword-parity (+--self-test), check-comment-mask-corpus, check-declaration-mirrors (+--self-test), check-position-name-fold-loaders (+--self-test), check-scripts-symbol-anchors (+--self-test), check-self-test-wired (+--self-test), check-self-test-workflow-commands (+--self-test), check-whole-set-label-write (+--self-test), pm/bare-root-worklist --self-test, pnpm check:agent-test-spelling, bash32-floor, cli-command-ids, cross-package-test-inputs, driver-memory-census, entry-guard, gitlink-declared, nul-bytes, parse-guard, pm-dispatch-gates (detached per header: '✓ dispatch-gates self-test (fast tier): 1913 cases pass; 6 slow section(s) deferred', BATTERY_EXIT=0, 406.6s), pm-expected-skips, pnpm-filter-targets, pnpm-flag-spellings, ratchet-remedy-authority, refd-timer-probe, watch-hint-literal. --ran: '✓ dispatch-gates --ran: 33 derived famil(ies) accounted for — 33 run, 0 NOT-MEASURED (a DERIVED zero — all 33 recorded an exit code and none of them is 3).' Rule ⑤ (suites naming this file): pnpm check:pm-half-states EXIT=0 (4916 cases); dispatch-gates.data.mjs / dispatch-gates.self-test.mjs pins inside the battery (marker stays on line 133); node scripts/measure-self-test-floor.mjs EXIT=1 refusing its census over check-prior-rulings.mjs + dispatch-gates.mjs, identical on base 3fbef41 (pre-existing, untouched files) — NOT MEASURED for this file, its --probe row not run because the census refuses first. CI on PR #24892 head fbd5fa3 at report time (single read, no polling): 28 check-runs — success 2 (Check Changeset, No other open PR may claim the same issue), skipped 7 (18.x upgrade harness (opt-in) x2, Auto Label, Check Changeset, Check PR Size, Packed-tarball smoke (opt-in) x2 — all rostered names), in_progress 19 (incl. every Lint · lane, every Type Check · lane, Governed Surface Queue Guard, filter, Check Documentation Links); required Lint & Repo Gates / TypeScript Type Check / Test Core / Build Core: in_progress or not yet created.",
"mcp_calls": "0",
"api_writes": "2 — REST relay dispatches POST /repos/objectstack-ai/objectstack/dispatches (HTTP 204 each), executed by objectstack-fleet[bot]: stroke 1 (request fw-20261020T132101Z-32fbae, run 39429636268 success) = pr_create: POST /repos/objectstack-ai/objectstack/pulls (draft) + POST /repos//issues/24892/assignees [huangyiirene] + POST /repos//issues/24892/labels [skip-changeset], read back: body 8270 bytes sent = 8270 stored, identical, one footer; draft true; labels [skip-changeset]; assignees [huangyiirene]. Stroke 2 = comment: POST /repos//issues/24849/comments (this report). Plus git push x3 to the branch (not REST). Reads: single-card REST GETs (issue 24849 + comments, PR 24771, its head commit/check-runs/check-suite, contents listings, PR 24892).",
"open_questions": [],
"out_of_scope_findings": [
"carrier: none · noted in PR #24892 Acceptance notes, not filed — node scripts/measure-self-test-floor.mjs refuses its census (exit 1: '2 file(s) carry a landed handshake spelling the recogniser reads as none: scripts/pm/check-prior-rulings.mjs, scripts/pm/dispatch-gates.mjs'), identical on base 3fbef41; a measurement tool with no CI wiring and no public door (drift, not class a/b/c). Dedupe words: measure-self-test-floor census refuses · classifyHandshake landed spelling · LANDED_HANDSHAKE_NAMES"
],
"line_budget": "scripts/pm/check-expected-skips.mjs 1499 -> 1727 lines (+247 / -19, 266 changed lines, 1 file; under the 3000 human-merge threshold). No skills/** or ratcheted ledger touched, so no line ratchet applies.",
"deviations": [
"Route vs the card's literal Done-when wording: the comparison is NOT checkout HEAD vs local origin/main nor vs the PR's base.sha; it is the roster's working-tree source blob ids vs the base branch's live tip via the contents API. Both literal candidates were measured to fail (behind-count refuses ~98% of merges and lies green on shallow clones; local origin/main is fetch-stale; PR #24771's base.sha 580fd4e would have compared the stale 6b22246 as ahead). Done-when's intent — behind ⇒ refuse naming the gap and the re-pin; self-test covers behind and at-tip; offline paths unchanged; no workflow edit, no roster row — holds.",
"Refusal was chosen over printing the gap (the card left it to the script maintainer); folded into exit 3, no new exit code — reason in the PR body.",
"A different board (PM_SWEEP_REPO pointing elsewhere) is now refused (exit 3) instead of judged against objectstack's roster — a direct consequence of the comparison; the header paragraph that described cross-board judging was rewritten; matches SKILL.md's '名单 check-expected-skips.mjs 只判 objectstack'.",
"test:repo (vitest repo project) not run locally: the derived list does not name it and the diff touches no test of that project; it belongs to CI's Test Core. No ① build / ② package tests: the diff touches no package.",
"HTTPS_PROXY's value (a loopback proxy URL) appears in transcript output: printed by the tool's own pre-existing re-exec line (the imported proxyRearmPlan reason) on each live run; no token or secret value was printed; presence checks used the non-printing form.",
"Commit trailers use AGENTS.md's model-free pair (Claude-Session + Co-authored-by: Claude), not the harness reminder's model-named form, per that reminder's own priority clause."
],
"files_changed": [
"scripts/pm/check-expected-skips.mjs"
]
}
Generated by Claude Code
objectstack-fleet commentedon Oct 20, 2026
ACCEPT → PR #24892 comment 6292161423 ·
domain:skillsseat 1 ·session_01M82feuetLzrCiBz41BJb7L· 2026-10-20T13:37ZGenerated by Claude Code