ui publish-check texts drift after #26860: release.yml publish-step comment, the dead ui changeset:publish alias and its wiring-test comment

Filing gate ① a stale text / dead alias, measured by the #26860 card thread (PR #27014, os-dev report 6458219555 on #26860, out_of_scope_findings 1–2); filed by the domain:ui#5 seat (session_01JQwxbXGpVJHZTXXerMoY4C, seat post #26945). #26860 asked for item 1 ("refresh … release.yml's comment over the publish step"); the seat kept it out of PR #27014 because .github/workflows/** is governed (Tier H) and the rest of that PR is not.

finding · tooling · reach: maintainers reading the release workflow and the ui/ scripts; no runtime effect.

What is stale

.github/workflows/release.yml: the comment over the publish step still describes the publish checks as they were before PR ci(ui-merge): port the four objectui gates that guard published @object-ui packages into root CI #26854 / PR fix(ui/scripts): check:node-esm-load and check:published-dist build modes run the release's build #27014 (the card text: it describes the absence of --no-build). Governed path: Tier H, director contract review before any approver.
ui/package.json changeset:publish runs check-published-dist-tooling.mjs with no --no-build, and nothing invokes it: the release publishes through scripts/release-publish.sh. Either remove the alias or make it say what it is for.
ui/scripts/__tests__/check-published-dist-tooling.test.ts: the wiring test's comment still calls changeset:publish "the command changesets/action runs".

Items 2–3 are ui/ (not governed); item 1 is governed. Triage decides whether they ride one PR or two.

Dedupe

Open issues read by REST for changeset:publish, release.yml publish comment: none beyond #26860 (closing with PR #27014) and #26717.

Triage: `domain:skills`, p3, `area:devpath` · one PR · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-28T01:58Z

⛔ Not a claim.

Read fresh. There are three stale texts after ui(scripts): the maintainer-run build modes of check:node-esm-load and check:published-dist are broken (exit 254; builds the docs app) #26860: the publish-step comment in release.yml, the dead ui changeset:publish alias, and its wiring-test comment. None has a runtime effect, so p3.
One PR, lane domain:skills. Item 1 is governed (.github/workflows/**, Tier H), so any PR carrying it is Tier H anyway. Items 2 and 3 are two lines that describe the same publish path, so splitting would only add a second review of the same fact.
⛔ This is a text edit to release.yml. No seat dispatches, re-runs or tests the release workflow by running it (Prime Directive 15). Item 2: remove the alias, or make it say what it is for, and say in the PR which one and why.
finding removed, pm:queue added.

Generated by Claude Code

Claim: PM loop round 1 (takeover shift, wave 3) · 2026-10-28T05:54Z
Session: session_01QSeKYmhU5KEj1nukYhp1Tg (seat; implementation by an os-dev subagent of this session, mode:subagent — same session ID, identity by branch)
Account: os-justin (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-27017-ui-publish-check-texts
Worktree: objectstack-issue-27017
Domain: domain:skills
Seat: domain:skills#2 (seat post #19287)
File surface: the three places the card names, text only — .github/workflows/release.yml (the comment over the publish step, from L1947 on origin/main 5fe0b967, IF it misdescribes what scripts/release-publish.sh does today; the dev measures: L67–74 run ui/scripts/check-published-dist-tooling.mjs --no-build before changeset publish), ui/package.json L116 (the changeset:publish alias nothing invokes — remove it, or make it say what it is for, and say which in the PR), ui/scripts/__tests__/check-published-dist-tooling.test.ts (the wiring test that pins that alias 4 times and whose L506 comment still says changesets/action runs it — follows the alias decision). ⛔ No behaviour change to the release path, ⛔ the release workflow is never dispatched or run (PD 15). Tier H (release.yml): one draft PR as triage 6458605256 ordered, governed approval. (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: TIER_DEFAULT (reason: dispatch-gates --tier --repo objectstack-ai/objectstack .github/workflows/release.yml ui/package.json ui/scripts/__tests__/check-published-dist-tooling.test.ts at fbc16563 prints no path-derived mandate; text and a dead alias, mechanical), effort: medium
Clause-②: no
Metadata-protocol: none
Responsibility: n/a — not a defect card (stale text / dead alias, filing gate ①, no runtime effect; triage 6458605256 p3)
Thread-read: 6458605256
Premise: items 2–3 hold on origin/main 5fe0b967 (read in this act, 2026-10-28T05:54Z): ui/package.json L116 "changeset:publish": "node scripts/check-published-dist-tooling.mjs && changeset publish", the wiring test names changeset:publish on 4 lines, and release.yml runs pnpm run release (= build + console:build + scripts/release-publish.sh), never changesets/action nor the ui alias. Item 1 is UNVERIFIED by the seat: the current release.yml comment from L1947 describes pnpm run release and names neither the dist-tooling check nor --no-build; whether that is the stale text the card means, or already accurate, is the dev's first measurement — a premise_still_valid: false on item 1 with items 2–3 done is a good outcome.
Serial constraints cleared: no open PR touches release.yml, ui/package.json or the wiring test (file scan of the open PRs in this act, 2026-10-28T05:54Z: none touches them); ui/** files are the ui lane's tree — triage assigned the whole card to this lane as one PR (6458605256), the seat declares that cross-lane surface here.
Resume-from: none

os-dev-report

{
  "issue": 27017,
  "status": "done",
  "branch": "claude/issue-27017-ui-publish-check-texts",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/27131",
  "head": "7076bbdd",
  "session": "session_01QSeKYmhU5KEj1nukYhp1Tg",
  "premise_still_valid": true,
  "premise_by_item": {
    "1_release_yml_comment": "true. Measured on origin/main a27f889a: the comment over the publish step (L1947-1951) said scripts/release-publish.sh 'is changeset publish followed by ONE atomic git push origin --tags'. Since daf52bcc (#26854) the script first runs node ui/scripts/check-published-dist-tooling.mjs --no-build (L73) and exits 1 before changeset publish when it is red. Since 4b7dbc44 (#26650) it also records npm's per-package answers for release-verify-npm.mjs. The comment left out the check the script refuses on, so it was corrected in place. This is a comment-only change: no step, run: line or condition moved. The PR is Tier H.",
    "2_ui_changeset_publish_alias": "true. git grep -n 'changeset:publish' outside ui/package.json found the wiring test (4 lines), a past-tense history comment in scripts/release-publish.sh L65, and released ui/packages/*/CHANGELOG.md entries (release-owned, not edited). No workflow or script invokes it. REMOVED rather than re-described: JSON has no comment spelling, and the alias was a second publish door that scanned a different build than the release's (it ran the scan without --no-build).",
    "3_wiring_test": "true. The comment at L506 ('changeset:publish is the command changesets/action runs') was false, because release.yml runs pnpm run release. The pin moved to the live path with the invariant kept."
  },
  "summary": "The release.yml comment over the publish step now lists what scripts/release-publish.sh runs, in order: (1) the @object-ui dist-tooling scan with --no-build, which refuses the publish; (2) changeset publish, with npm's answers recorded for release-verify-npm.mjs; (3) one atomic tag push. The dead ui changeset:publish alias is removed. Its wiring test now pins the live path: the root release script ends in bash scripts/release-publish.sh, and that script runs `if ! node ui/scripts/check-published-dist-tooling.mjs --no-build; then` before its changeset publish line (comment lines are stripped before the check). The whole card rides one draft PR in lane domain:skills as triage ordered, and the ui/** files are declared as cross-lane in the body.",
  "tests": "Edited file at 12551505 and again at 7076bbdd: pnpm --dir ui exec vitest run --maxWorkers=2 scripts/__tests__/check-published-dist-tooling.test.ts, under the lock -> Test Files 1 passed (1), Tests 35 passed (35), VERDICT command-exit 0. Reverse check, done on the REAL files because the test reads fixed paths. The fix was committed first (12551505). Each leg used scripts/ablation-replace.mjs (anchor hit 1 -> 0, blob moved), a git checkout HEAD trap, and an explicit hash-object == HEAD-blob check afterwards: (a) '--no-build' dropped from the gate line in scripts/release-publish.sh -> 'runs on the publish path' FAILED (expected ... to match /^if ! node ui\\/scripts\\/check.../), exit 1, restored blob 6d228a57 == HEAD; (b) root package.json release '&& bash scripts/release-publish.sh' -> '&& changeset publish' -> FAILED (expected false to be true), exit 1, restored a0c7eb67 == HEAD; (c) 'unset npm_config_globalconfig ...' line replaced by 'changeset publish' so a publish line precedes the gate -> FAILED (expected 43 to be less than 19), exit 1, restored. A first attempt at (c) was refused by ablation-replace (the anchor was contained in the replacement, so the count did not drop): nothing was written, and it was re-run with a different anchor. Green again after restore: 35/35, exit 0. The direction was the expected one (red) on all three legs. No dist rebuild was needed: the test reads source text files, and no package exports are involved.",
  "gates": "Head 7076bbdd (origin/main merged twice; families unchanged between 9849e9ec and 7076bbdd). node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 62 commands. All 62 ran on 7076bbdd with exit captured before any pipe: 60 direct/pnpm gates exit 0 (incl. check:required-contexts, check:workflow-status-functions, check:pm-expected-skips, check:workflow-step-name-quoting, check:cross-package-test-inputs, check:nul-bytes, check:published-files, check:type-check-coverage/debt, release-verify-npm/release-dist-tags --self-test); pnpm check:pm-dispatch-gates exit 0 (fast tier 1935 cases); pnpm --dir ui exec vitest run scripts/ exit 0 (175 files passed, 2 skipped; 5019 tests). --ran with ':: exit N' records: '62 derived, 62 run, 0 NOT-MEASURED (a DERIVED zero)', exit 0. That --ran pass warned the tree is 8 commits behind origin/main, with lint.yml changed; no conflict, so main was not merged a third time (os-dev: merge after open only on conflict or main-caused red). Outside the derivation: ui eslint on the edited test, exit 0. On the fresh worktree the ui scripts suite first needed @objectstack/spec, @objectstack/formula and @objectstack/client built: on 9849e9ec, before the builds, 3 files failed on missing dist (unit-registry-absence-collision, date-only-zone-pins-10183, check-vi-mock-inherit), and none of them reads a changed file. No check-release-* family was derived. PR check-runs: not awaited, in_progress at report time.",
  "files_changed": [
    ".github/workflows/release.yml",
    "ui/package.json",
    "ui/scripts/__tests__/check-published-dist-tooling.test.ts"
  ],
  "line_budget": "38 changed lines (+27/-11), 3 files",
  "dogfood_battery": "n/a: no ! in the PR title, and Clause-② no without a (narrowing) arm",
  "mcp_calls": "0 GitHub MCP calls",
  "api_writes": "3 classes, as budgeted: (1) git push of the branch: the empty probe, commit 12551505, then merge commits 9849e9ec and 7076bbdd, all fast-forward with no force; (2) relay pr_create via POST /repos/objectstack-ai/objectstack/dispatches (fleet-write run 40650389817, success) -> PR #27131, draft, assignee os-justin, label skip-changeset, body read back 5787/5787 bytes identical, labels read back by REST GET: skip-changeset; (3) this os-dev-report comment via post-stamped -> POST /repos/objectstack-ai/objectstack/issues/27017/comments",
  "deviations": [
    "Base moved: the order named origin/main 5fe0b967. The worktree was cut from a27f889a, and origin/main was merged in twice (final head 7076bbdd). The release.yml comment text at L1947 was unchanged across that range.",
    "Commit trailer: used the model-free pair that AGENTS.md requires (Co-authored-by: Claude), not the harness reminder's model-named Co-Authored-By. PR footer: used the os-dev session-URL form, not the harness '🤖 Generated with' line. In both cases the role file / AGENTS.md wins.",
    "The first pnpm check:pm-dispatch-gates run had no exit capture. It was re-run with capture on 9849e9ec (EXIT 0) and again on 7076bbdd (EXIT 0).",
    "Built @objectstack/spec, @objectstack/formula... and @objectstack/client... closures, outside ① (the diff touches no package), so the derived ui scripts suite could run on a fresh worktree."
  ],
  "open_questions": [],
  "out_of_scope_findings": [
    "carrier: none · noted, not filed. scripts/release-publish.sh header L25-27 still says it is 'Run as the publish: command of changesets/action', and release.yml says changesets/action is not used. This is stale text with no runtime effect, so it fits none of classes a/b/c. Recorded in PR #27131 Acceptance notes."
  ]
}

Generated by Claude Code

ACCEPT → PR #27131 comment 6463573514 · domain:skills#2 · 2026-10-28T07:47Z — ⛔ not a claim. Contract review (CONTRACT_REVIEW_TIER) PASS → PR comment 6463562292 on head 7076bbdd. Tier H (.github/workflows/release.yml): os-zhuang APPROVED 5679210902 on this head (2026-10-28T07:31Z); who also readied and enqueued the PR himself (2026-10-28T07:43Z); the seat flipped nothing, follows it to MERGED and strips pm:dispatched when Fixes #27017 closes the card. Acceptance notes: the stale changesets/action header in scripts/release-publish.sh L25–27 — noted, not filed (no class a/b/c reach).

Landed — PR #27131 merged; card closed by `Fixes #27017`

probe	reading (2026-10-28T08:21Z, `domain:skills#2`)
`git log -1 --grep='#27131' origin/main` after `git fetch origin main`	`4db45988` `ci(release),ui: the publish-check texts say what the release does today — the dead ui changeset:publish alias goes, its wiring test pins the live path (#27131)` — on `origin/main`; `GET pulls/27131` → `merged: true`, `merge_commit_sha 4db45988`, merged by the queue entry `os-zhuang` made (2026-10-28T08:08Z)
`GET issues/27017`	`state: closed`, `state_reason: completed` (2026-10-28T08:08Z), closed by the PR body's `Fixes #27017` (the only closing keyword in the body — no other card was closed by it)

State change: pm:dispatched removed from this card in the same act (label-write, read back: tooling, domain:skills, priority:p3, area:devpath remain as ownership). Acceptance note carried from the PR, not filed: scripts/release-publish.sh L25–27 still says the script runs as the publish: command of changesets/action (stale header prose, zero runtime effect).

Generated by Claude Code

Navigation Menu

Repository navigation

ui publish-check texts drift after #26860: release.yml publish-step comment, the dead ui `changeset:publish` alias and its wiring-test comment #27017

What is stale

Dedupe

Triage: `domain:skills`, p3, `area:devpath` · one PR · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-28T01:58Z

Landed — PR #27131 merged; card closed by `Fixes #27017`

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

ui publish-check texts drift after #26860: release.yml publish-step comment, the dead ui changeset:publish alias and its wiring-test comment #27017

Description

What is stale

Dedupe

Activity

objectstack-fleet commented on Oct 28, 2026

Triage: domain:skills, p3, area:devpath · one PR · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-28T01:58Z

objectstack-fleet commented on Oct 28, 2026

objectstack-fleet commented on Oct 28, 2026

objectstack-fleet commented on Oct 28, 2026

objectstack-fleet commented on Oct 28, 2026

Landed — PR #27131 merged; card closed by Fixes #27017

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

ui publish-check texts drift after #26860: release.yml publish-step comment, the dead ui `changeset:publish` alias and its wiring-test comment #27017

Triage: `domain:skills`, p3, `area:devpath` · one PR · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-28T01:58Z

Landed — PR #27131 merged; card closed by `Fixes #27017`