ci: a scheduled workflow regenerates scripts/test-shard-timings.json from the newest fully green main run and opens the PR (maintainer-directed, measurement loop)

[claude]

Filed by the skills lane seat (session session_019RfFHiRCSs3JXLK4cwcfox, os-steve) on the maintainer's direction, 2026-09-07T02:5xZ, and dispatched by the same seat on that direction. Surface owner stays domain:devx.

Authority (maintainer, verbatim, live PM chat 2026-09-07): 「同意你的建议，你负责执行派发所有可行的优化」, on the seat's long-term plan for the Test Core shards (measurement loop → capacity rule → growth constraint). This card is the refresh half of the measurement loop; #16173 carries the one-time refresh and the drift alarm's wiring.

What

A scheduled workflow regenerates scripts/test-shard-timings.json from the six test-core-run-summary-N-of-6 artifacts of the most recent complete, fully green, full-scale push run on main, and opens a pull request with the result, so the balancing input never again depends on someone remembering.

Measured

• The file is generated only by hand (provenance.refresh); it went 13 days without a refresh while ~700 test files were added, and the shard it mis-weighted was killed twelve times in a day (CI: the shard-timings file is stale for the CLI package — 672s predicted vs 28m46s measured against a 30-minute timeout, so Test Core shard 1/6 is one slow run from being killed on any PR touching the CLI #16173).
• The inputs already exist on every run: each shard publishes its turbo summary as an artifact (retention 1 day); scripts/measure-test-shard-timings.mjs takes any number of summaries and merges by median.
• Push runs on main are cancelled by cancel-in-progress when merges come quickly, so the workflow must select a run whose six shards all completed success (uncensored), not merely the newest.

Ruling

1. New workflow .github/workflows/shard-timings-refresh.yml: schedule weekly (Monday 05:30Z) plus workflow_dispatch. Steps: list the CI workflow's push runs on main through the API with the workflow token, pick the newest whose six Test Core (N/6) jobs all concluded success and whose artifacts are still retained, download the six summaries, run the measure script with --out scripts/test-shard-timings.json, run partition-test-shards.mjs --self-test and update its balancing pins the way the script's own refresh procedure prescribes, and open a PR on a bot branch with skip-changeset, the run id and the before/after per-shard prediction in the body. No PR when the file is byte-identical.
2. The workflow never hand-edits the JSON and never touches timeout-minutes or the shard count; it is the documented refresh command on a timer.
3. Read cut-rc.yml and pr-automation.yml for the house idiom of a workflow that pushes a branch and opens a PR; follow it.

Acceptance

• workflow_dispatch once on the PR branch: the PR it opens (or its "byte-identical, no PR" log line) is quoted; the seat reads it.
• The workflow's own gates green (dispatch-gates --commands derives them).

Refs #16173.

---

Generated by Claude Code

[claude]

Claim: PM loop round 9 — flight C2a, maintainer-directed (「同意你的建议，你负责执行派发所有可行的优化」, 2026-09-07): the scheduled shard-timings refresh workflow, with #16473 folded in as its prerequisite — flight C1 (#16173) measured at 03:2xZ that the artifact host (productionresultssa*.blob.core.windows.net) is denied to every agent container by egress policy, so the documented refresh cannot run from a dev worktree; the refresh therefore runs IN-RUNNER, where .turbo/runs/*.json already sits, and this workflow's first workflow_dispatch is the refresh itself — the PR it opens is #16173's deliverable. Rider #16473: buildDataset in scripts/measure-test-shard-timings.mjs bypasses its median rule for a file-sharded package (last run wins; slices from different runs summed into a weight no run measured) — fixed first, with self-test cases, so the refresh's @objectstack/cli weight is a median like every other package's.
Session: session_019RfFHiRCSs3JXLK4cwcfox
Branch: claude/issue-16464-shard-timings-refresh-workflow
Worktree: objectstack-issue-16464
Domain: domain:devx (dispatched by the skills seat on the maintainer's direction)
File surface: .github/workflows/shard-timings-refresh.yml (new), scripts/measure-test-shard-timings.mjs (the slice ledger's merge rule + self-test), one helper under scripts/ci/ if the pin update needs one; scripts/test-shard-timings.json is written only by the workflow run, never by hand. ⛔ ci.yml untouched; no timeout-minutes, matrix or bound change.
Container & model: S, mode:subagent, opus — dispatch-gates.mjs --tier: no path-derived mandate.
Clause-②: no
Serial: no open PR touches the measure script or the JSON (C1 produced no PR); flights V and X hold ci.yml, which this card does not touch. Not governed ⇒ in-seat review, then the seat flips ready and arms SQUASH auto-merge; skip-changeset. The PR carries Fixes #16464 and Fixes #16473; #16173 and #16222 are Refs (they close when the workflow's first refresh PR merges).

---

Generated by Claude Code

[claude]

File-surface supplement to the Claim: above (not a second claim; branch, session and worktree are unchanged). One file is added to the declared surface, under the bounded in-place exemption, because the card's ruled deliverable cannot exist without it.

Added: scripts/check-self-test-wired.mjs — its invocation extractor had a left boundary but none on its extension alternation, and js is a prefix of json. The first workflow line naming a .json under scripts/ therefore matched only as far as the .js and minted a phantom key with no file behind it:

scripts/test-shard-timings.json   ->  keyed as  scripts/test-shard-timings.js

That is exactly the phantom the gate's left boundary was added to abolish, arriving from the other end, and it is silent in both directions — the real file is audited by nothing and the phantom key reconciles against no carrier.

Why this is not scope creep: ruling 2 requires the workflow to invoke the generator with --out scripts/test-shard-timings.json, and rulings 3 and 4 require it to byte-compare and commit that same path. Writing the path is the card's content, so there is no version of this deliverable that does not trip the gate. It was invisible until now only because no workflow had yet named a .json under scripts/.

The four conditions, each checked rather than asserted:

1. Same class — a path-matching rule that reads a name it was not given, which is the class this card's own generator defect ([finding] measure-test-shard-timings: a file-sharded package bypasses the median merge rule — one run wins, and slices from different runs are summed as one measurement #16473) belongs to.
2. Mechanical, with the correct form already fixed by the file — a negative lookahead rejecting a following word character or hyphen is the spelling that file already applies to the self-test flag two groups along, and its header states the exact-equality principle the extension had not been given.
3. No other claim holds it — all 33 open PRs were read for their file lists at 04:0xZ; only ci: the merge queue's Test Core runs the affected package set; selection moves into a script with a self-test #16471 touches scripts/ci/, and only select-shard-packages.sh / .selftest.sh, which are different files from this card's select-shard-timings-run.mjs. Nothing open touches the gate.
4. Same gate family, no new verification surface — check-self-test-wired and its --self-test were already in this card's derived family before the edit.

Proven red-then-green rather than argued. With the boundary removed and the mutation confirmed on disk (anchor 1 to 0, git diff --numstat = 1 1), the production run stays GREEN at exit 0 while --self-test goes RED naming the new pin; restoring gives back the byte-identical blob (db95b1025a4ba4d6a310268161d903b6d44af716, equal to the HEAD blob) and the self-test returns to exit 0. Four cases pin it, control first, battery floor 4 to 8.

Declared surface is therefore: .github/workflows/shard-timings-refresh.yml (new), scripts/ci/select-shard-timings-run.mjs (new), scripts/measure-test-shard-timings.mjs, scripts/check-self-test-wired.mjs. ci.yml is untouched; scripts/test-shard-timings.json is written only by a workflow run, never by hand.

---

Generated by Claude Code