fix(ci): reformat 09-event-schema.md table to unbreak the Prettier lane - #13588

✅ Qwen Triage finished — view run. See the stage comments in this thread for the result.

✅ Qwen Triage 已完成 —— 查看运行。结果见本线程中的各阶段评论。

Thanks for the PR — and for digging into why the lane went red instead of just papering over it. The diagnosis in your description is accurate. The problem is that it has since been fixed underneath you.

Template looks good ✓

Problem: real when you filed it, no longer real on main. The violation did land — #13281 (merged 04:45Z today) put *genuinely queued* plus over-padded table cells into docs/developers/daemon/09-event-schema.md, and because that PR was docs-only it never ran a lint lane at all, so nothing caught it. A PR that does run the full lane goes red on exactly this: #13567's Lint & Static job fails at step Run Prettier, and its tree still carries the asterisk form. Your reasoning about the repo-wide --check . is correct.

But #13536 merged at 08:29:34Z — six minutes after you opened this at 08:23:32Z — and its own patch to this same file is the identical +4/-4 (*genuinely queued* → _genuinely queued_, same re-padding). I pulled both blobs and compared them directly: main's copy of docs/developers/daemon/09-event-schema.md is now byte-identical to your head — both md5 6d9f5fb8a87e67a4cbf9e4b69b8b366c, 369 lines / 47224 bytes. compare main...5826353a reports diverged (ahead 1, behind 1) with a +4/-4 file entry, but that is measured against the merge base; the net content delta against today's main is zero.

Direction: aligned — a gate that any merged commit can break for everyone is worth fixing fast, and fixing it with formatter output rather than by hand is the right instinct. Simply overtaken by events here.

Size: not applicable — one docs file, no core paths.

Approach: minimal and appropriate for the goal: four lines, formatter output only, no wording changes. Nothing I'd cut.

One thing beyond this PR that a maintainer should see: this is a four-way duplicate. #13580 (06:51Z), #13584 (07:27Z) and #13587 (08:01Z) are all still-open qwen-code-dev-bot PRs with the same +4/-4 on the same single file. All four were obsoleted by #13536 within the hour.

Risk: no elevated risk signals — no revert-correlated paths touched.

Flagging this before diving deeper, since the substantive finding is that there is nothing left to merge. Code review and CI evidence follow in the next comment.

— Qwen Code · qwen3.8-max-2026-09-02

_{Reviewed at 5826353a6d5906c7d73d7f0d6cb6f53d8271ab25 · re-run with @qwen-code /triage}

Independent proposal first (written before reading the diff): given "a merged docs commit broke the repo-wide Prettier lane", I would run the lockfile-pinned formatter over the offending file and commit exactly its output — nothing more, no hand-tuning, no accompanying cleanup. That is precisely what this PR does, so the approach matches my baseline. The one step I would add is checking whether the fix had already landed on main before opening — that is the step that would have caught this.

Code review: clean, no blockers, no AGENTS.md violations. The diff is formatter output and nothing else: Prettier's default markdown emphasis normalisation (*genuinely queued* → _genuinely queued_, which renders identically) plus one character of re-padding in the Trigger column, with the longest cell in that column (pending_prompt_completed) left untouched. No wording, semantics, or table-structure change.

One small inconsistency in the description, not a blocker: the summary says "Whitespace and table alignment only, no wording changes", while Risk & Scope correctly discloses that "emphasis-marker style" also moves. The Risk section is the accurate one — emphasis markers are markup, not whitespace.

Skipping both optional enrichments (sequence diagram, changed-files table): one file and four lines, so a diagram or a per-file map here would be noise rather than signal.

Test evidence — the PR's own CI, read through the API. No PR-derived code was executed.

The green lint lane on this PR is vacuous, and that matters because the description leans on it. Lint & Static (ubuntu-latest, Node 22.x) reports success, but this PR is docs-only, so the job skipped every real lane step — Install dependencies, Run ESLint, and Run Prettier (step 27) are all skipped. Nothing in this PR's CI ever invoked prettier --check. A green check that never ran the thing being fixed is not evidence for the fix.

Because this PR's own lane cannot substantiate the claim, I went and found CI that does — and it settles both directions:

• The problem was real. fix(core): resolve modelProviders-prefixed model selectors for subagents #13567 (fix(core), whose tree still holds the asterisk form) runs the full lane, and its Lint & Static job fails at Run Prettier. That is a genuine red gate caused by exactly the violation this PR targets. It also explains how the break landed undetected: docs(serve): fix stale permission-route note and document pending-prompt events #13281 was docs-only, so no lint lane ran on it at all — its merge commit carries no Lint & Static check.
• The fix is correct, and already on main. feat(managed-agent): H6a schedule and automation run record contract #13536's head ran the real lane on content byte-identical to this PR's — Install dependencies success, Run ESLint success, Run Prettier (step 27) success — with Prettier resolved from pnpm-lock.yaml at 3.6.1. So the reformatted table is confirmed Prettier-3.6.1-clean by an actual gate run, not by inference. I did not need to run a formatter myself (this run executes nothing), and that leaves no hole in the claim.

The npx [email protected] --experimental-cli --check . output pasted in the description is the author's local result, not independently re-run — but it is no longer load-bearing, since #13536's Run Prettier step covers the same ground with the same pinned version. The aside about a floating 3.9.9 restyling 703 files is flagged out-of-scope by the author and nothing in the merge decision rests on it.

Sandboxed lane: none applies, and I want to be explicit about that rather than silently skip it. This is a docs-only change with no runtime surface, so @qwen-code /verify (A/B behavioural proof against the base build) and @qwen-code /tmux (TUI capture) have nothing to settle — and /tmux is unavailable to this author regardless, since they hold read-only permission. The only open question was a formatter-version one, and a real lint lane answered it above; no sandbox would have added anything.

Real-scenario testing: N/A — unattended CI run, and nothing user-visible to drive.

— Qwen Code · qwen3.8-max-2026-09-02

_{Reviewed at 5826353a6d5906c7d73d7f0d6cb6f53d8271ab25 · re-run with @qwen-code /triage}

Confidence: 3/5 — the change itself is correct and was well-evidenced when filed, but #13536 landed the identical formatting six minutes later, so there is no delta left to merge and nothing for me to approve.

Stepping back: this is a good PR that arrived too late. The diagnosis was right, the fix was the minimal formatter output rather than a hand-edit, the description named the exact pinned version and even flagged the 3.6.1-vs-3.9.9 drift as out of scope instead of quietly widening the diff. Had main still held the asterisk form, I would have approved it without hesitation. My independent proposal was the same change, so there is no simpler path I can offer — the only thing I'd have done differently is check main first.

What stops this from being a clean approve is that approving it would attest to a change that no longer changes anything. main's copy of the file is byte-identical to this PR's head (md5 6d9f5fb8a87e67a4cbf9e4b69b8b366c on both), so merging adds a commit with a zero net diff. I'm also not going to request changes: the author did nothing wrong, and a blocking review state would be the wrong instrument for "you got raced by another merge."

So the recommendation is to close this as superseded by #13536. I'm not closing it myself — triage's only automated close path runs through a linked issue that was closed as completed by a merged PR, and this PR deliberately has no linked issue, so that path doesn't apply. @he-yufeng, if you agree, closing it from your side is the cleanest ending; if you think something here is not covered by #13536, say so and I'll re-run.

Two things a maintainer should see, neither of them this author's fault:

1. This is a four-way duplicate. fix(docs): apply Prettier formatting to the daemon event schema (#13573) #13580, fix(docs): apply Prettier formatting to the daemon event schema (#13581) #13584 and docs(serve): fix Prettier formatting in the daemon event schema (#13585) #13587 are all still open, all from qwen-code-dev-bot, all +4/-4 on this same single file, opened 06:51Z / 07:27Z / 08:01Z. All four are obsoleted by feat(managed-agent): H6a schedule and automation run record contract #13536 and want the same treatment. I evaluated this one on its own merits, but the cluster itself is the signal — the repo's own bot generated three copies of one markdown fix in under two hours.
2. The blind spot that caused it is still open. docs(serve): fix stale permission-route note and document pending-prompt events #13281 was docs-only, so it ran no lint lane at all, and a repo-wide prettier --check . violation reached main undetected. It only surfaced on the next non-docs PR (fix(core): resolve modelProviders-prefixed model selectors for subagents #13567), which then went red for something it didn't cause — and that red is what attracted the duplicate fixes. Check lint gate freshness was skipped on this PR too. Worth a look: either let the Prettier lane run on docs-only PRs, or add a post-merge format check on main, so a docs-only merge can't break the gate for everyone behind it. Raised as a question, not as a condition on this PR.

No approval posted in this run, and no approve-on-green marker — the verdict is defer, not approve. The deterministic maintainer resolver returned nothing here (QWEN_MAINTAINER_HANDLE unset, no area labels on the PR, no prior human reviews), so rather than guess an owner I'm routing this to @wenshao, who merged #13536 and therefore has the context on both the duplicate cluster and the docs-only lint gap. Flagging for a human call; nothing here blocks your other work.

— Qwen Code · qwen3.8-max-2026-09-02

_{Reviewed at 5826353a6d5906c7d73d7f0d6cb6f53d8271ab25 · re-run with @qwen-code /triage}

📄 Docs-only change — the automatic review ran at --effort medium (verified findings, no reverse audit; medium posts no inline comments). Outcome:

Review complete: pr-13588 — Comment, not posted (0 Critical, 0 Suggestion)

Reviewed head: 5826353a6d5906c7d73d7f0d6cb6f53d8271ab25. Full report in the workflow run. For a full high-effort review with inline comments, comment @qwen-code /review.

Closing: #13536 landed the identical Prettier reformat of 09-event-schema.md on main at 08:29Z, six minutes after this was filed, so there is no delta left to merge. The diagnosis still holds (the lane break was real at filing time); the fix just arrived elsewhere first.