Repository navigation
platform-readings: mergeable_state: unstable can come from a commit STATUS that the /check-runs enumeration cannot see #18521
Description
A one-line fact-table refinement, measured today. Not a defect in anyone's code — a gap between what mergeable_state: unstable means and what the reading the seats actually take can see.
What is recorded today
.claude/skills/pm-dispatch/references/platform-readings.md:27:
ready 翻转实测两序列
clean→blocked→clean与blocked→unstable→clean;unstable瞬态非失败。
True, and it saved a false alarm today. What it does not say is where the unstable comes from, and that matters because the landing pre-check's condition ③ ("every check green") is normally taken by enumerating GET /repos/{o}/{r}/commits/{sha}/check-runs and judging the latest run per check NAME.
The measurement
PR #18486, head 750aa5b0954a31434d6c6df428478fad8b6035ed, read 2026-09-16T18:54:56Z, after the ready flip and after the PR entered the merge queue:
| reading | value |
|---|---|
GET /commits/{sha}/check-runs, latest run per NAME |
35 names — 29 success, 6 skipped, 0 failed, 0 in progress |
GET /pulls/18486 |
mergeable: true, mergeable_state: **unstable** |
GET /commits/{sha}/status |
state: **pending**, contexts [('Vercel', 'pending')] |
⇒ the unstable is produced by a commit STATUS, which is a different API surface from check-runs and is invisible to the /check-runs enumeration. Here it is a third-party deploy preview (Vercel), it gates nothing, and the PR sat in the queue with it pending — so the existing line's verdict ("not a failure") holds. But a seat reading only check-runs cannot tell that story; it sees a clean roster and an unexplained unstable, and the honest options are to guess or to look somewhere it was never told to look.
Suggested line (one line, to sit beside :27)
unstable 的来源可能是 commit STATUS(如 Vercel 预览)而非 check-run:/check-runs 枚举看不见它,判因走 GET /commits/{sha}/status 的 state 与 statuses[]。
Why it is worth a line rather than nothing
Condition ③ of the landing pre-check is one of exactly three things standing between a reviewed PR and main. A condition whose instrument cannot see one of the inputs to the state it is judging is the shape that produces a confident wrong reading — today it produced a correct one only because the existing :27 line happened to cover the symptom.
Dedupe words: mergeable_state unstable, commit status vs check-runs, Vercel pending context, landing pre-check condition 3, combined status endpoint.
Generated by Claude Code
Activity
os-justin commentedon Sep 17, 2026
Claim: PM loop round 1
Session:
session_01Gqi43smmqjJ5sUrhfoPeKuBranch:
claude/issue-18426-platform-readings-batchWorktree:
objectstack-issue-18426Domain:
domain:skillsSeat:
domain:skills#1File surface:
.claude/skills/pm-dispatch/references/platform-readings.mdonly (466 / 466, ≤120 B per row) — a SIBLING card ADDED to the five-card batch already in flight on this branch (carrier #18426, claims 5716076148 / 5716077615 / 5716078793 / 5716079980 / 5716081402), on triage's 「强烈建议同派」: its reading —mergeable_state: unstablecan come from a commit STATUS the/check-runsenumeration cannot see — landing pre-check ③ reads BOTH/check-runsand/status(or the combined status) before calling a head green — adopted or rejected on its own with the reason on the PR; adopted ⇒ its ownFixes #18521line, rejected ⇒Part of #18521and the seat re-queues it; ⛔ keep off :10–:12 / :431 (#18469 PR-A) and :212 (PR #18666); FACT layer ⇒ lands on the seat's in-seat review of record, ⛔ no four-pieceContainer & model:
L(受管事实表批次追加一卡),mode:subagent,model: claude-opus-5— the batch's tier reading at 2026-09-17T14:27Z (no path-derived mandate; default tier) holds for the same fileClause-②: no
Thread-read: 5715654046
Serial constraints cleared: same branch as the batch — one PR; the seat re-scoped the running dev by message at 2026-09-17T14:43Z;
platform-readings.mdopen-PR overlap unchanged (PR #18666 :212 only); shared checkout atorigin/main6dfa3ea77. Total order: a p2 Task of the 14:02Z–14:03Z routing, folded because it shares the one file at zero headroom.Ruling-ref: 5715654046
Generated by Claude Code
os-justin commentedon Sep 17, 2026
Batched onto PR #18689 (draft, governed FACT layer) — this card's reading is ADOPTED as one row at
platform-readings.md:28(115 B), sitting directly beside the:27row it refines (「unstable瞬态非失败」) and not contradicting it, paid by compressing the two CI-log reading rows at:303. Verdicts, payments and gate exit codes are on the PR.Generated by Claude Code
os-justin commentedon Sep 17, 2026
LANDED — PR #18689 (the seven-card platform-readings batch, carrier #18426) merged through the queue at 2026-09-17T15:59:40Z (
merged_at), squash72dd95fa5(single parente75cc3c2f, an ancestor oforigin/main;platform-readings.md466 / 466 onmain, +17 / −17). Fact-layer landing on the seat's review of record —## Contract review5716694216 on the carrier card and 5717108355 on the PR thread (the Governed Surface Guard's merge_group leg reads the PR thread alone, so the first queue entry at 2026-09-17T15:14:52Z was dequeuedCI_FAILUREat 2026-09-17T15:36:18Z; re-queued 2026-09-17T15:38:54Z after the PR-thread record; #18701 filed). Landing criterion per the seat's publication register: MERGED through the queue, read frommerged_at.This card's row on
mainat :28: 「-unstable可源自 check-runs 看不见的 commit STATUS(如Vercel)⇒ ③ 另读/commits/{sha}/status。」. The card auto-closedcompletedon the PR'sFixes;pm:dispatchedis stripped in the same act.Generated by Claude Code
docs(pm-skills): adopt eight measured platform readings into the gove…