[objectstack-fleet]

Requested by Steve · project thread
Fixes #25842
Fixes #25847
Fixes #25848
Part of #25849 (4 of its 5 findings; the write-feedback acceptance[1] reading waits on a maintainer ruling, so the card stays open)
Fixes #25850
Clause-②: no
Metadata-protocol: none

Checklist-accuracy findings from the 18.2.0 release sweep, applied to five ledger files, one commit per card. Wave anchor: #25807 (it stays open; this PR only links it). Docs only: nothing under packages/, ui/ or any governed surface moves, so the PR carries skip-changeset.

Every JSON edit revises the item text, bumps its revision, and appends a history row that names the run record the finding came from (QA runs #25815 release-18.2.0-tier1 and #25817 release-18.2.0-manual-set, both at 6679644a; #25812 and #25823 for RUNNER). Each finding was re-read against origin/main at cc378e93 first, with objectui paths read under ui/.

Per card

#25842, areas/i18n.json (9 findings, all applied, 5 items).

• i18n.flow-copy-translation-and-host-override: the ref adds the three packages/cli/test/i18n-flow-*-coverage tests; the verify for the extract clause says they prove extractTranslations, not the command's file write; the validate clause says translation.test.ts pins the rule and its remedy while the exit code stays the run's.
• i18n.surface-matrix: the ref says the two objectui pins cover the confirm copy and the successMessage handed to the toast, under stubbed action, confirm and toast handlers (no round-trip), so the network clause stays the run's. It also says acceptance 6, 7 and 9 have no pin: the anonymous-deny dogfood's member 200 control asserts none of the locale set, the group vocabulary or the zz-ZZ shape.
• i18n.strict-translation-key-rejection: the steps name a snake_case item (the conventional name zh-CN is refused first with 400 INVALID_REQUEST), and the clause scores 422 INVALID_METADATA. A3 cites ADR-0049 and spec 17.0.0, as the live text does. A5 reads the singular label on the record-page badge and the New dialog, and pluralLabel on the list header and nav.
• i18n.notification-localized-and-clears: the steps and the clause read the server's unreadCount (GET /api/v1/notifications), never the header badge, which also counts pending approvals.

#25847, areas/search.json (6 findings, all applied, 4 items).

• search.cross-field-object-search: the pin proves the name-hit control by substring, not by naming Acme Retail.
• search.field-scoped-narrowing: the ref is scoped to acceptance 0. The pin has no $searchFields refusal case and drives only the POST spelling, so acceptance 1 to 3 and both negatives are the run's. The route-handler INVALID_FIELD evidence in packages/rest is named as corroboration.
• search.global-search-skips-unreadable: the dogfood pin runs on planted skipsearch_* objects, not the showcase objects, and its unscoped case asserts no totalObjects.
• search.pinyin-flag-both-sides: the run's own paths must not spell pinyin (they polluted the zero-hit grep), and the 6679644 counts are recorded (53 on, 52 off).

#25848, areas/studio-authoring.json (9 findings, all applied). Each automated.ref now says which clauses its pins score: jsdom and unit pins that stop above HTTP score only the test-oracle clause. Affected items are expression-editors, meta-save-preconditions-and-draft-rows, metadata-diagnostics-sweep (its pin's [#8924] describe is named as protocol-layer evidence for acceptance 3 and negative 0), object-designer-roundtrip (F4 is corroboration under a different fixture), org-override-registry-gate and save-door-runs-authoring-rules. In permission-matrix-editor-ux, the ref adds PermissionMatrixEditor.artifactTier.test.tsx for the artifact gate. In record-page-roundtrip, the ref names the run's objectui pin b41ed7f68956 beside cf153eb2e535.

#25849, areas/ux-conventions.json (5 findings: 4 applied, 1 answered). write-feedback:

• Import Undo works in create and update modes, so the Undo clause is scored on the import instead of blocked(fixture).
• Discard draft is an in-app dialog.
• The approvals route has no drawer: a request opens on its record page.
• The member's refusal row is Mark Done, because the console offers a member no inline edit, selection or drag.
• Not applied: the run asks for a ruling on object-only success strings ("Task created successfully", "Settings saved", Flow "Reassign Task" completed) under acceptance 1. That is a maintainer decision, so the clause is unchanged and the question is raised in the report on qa(checklist): i18n — 9 ledger edits from the 18.2.0 release sweep (#25807) #25842.

#25850, RUNNER.md (3 findings, all applied; re-read on top of #26052).

• Walled boot: examples/app-showcase/package.json declares @objectstack/organizations, so a walled run boots from the tree with three variables and no scratch copy.
• Playwright: /opt/pw-browsers holds chromium 1243, matching playwright-core 1.63.0, and the alias resolves there (re-read in this container at cc378e93). The entry now describes a mismatch as the case to check for.
• OS_PORT: os dev serves the prebuilt dist/objectstack.json, so the connector metadata's baseUrl can carry the build-time value while the live dispatch reaches the exported port. The Check now reads the self-ping flow's run.
• Walled personas: after feat(plugin-auth): multi-org Create User joins the administrator's current organization #25626, Create User joins the operator's current organization, so the operator creates an organization and runs set-active before minting personas.

Retired objectui-pin wording in the same files (#26035 left the area JSON to these cards). .objectui-sha, objectui:refresh and "not runnable from this checkout" are rewritten to ui/ paths and pnpm console:build. That covers i18n.studio-follows-app-locale, studio-authoring.first-run-loop, packaged-automation-studio-lock, packaged-display-class-direct-edit, doc-on-app-menu, picklist-bound-select, the four "AUTOMATION IS PINNED IN THE objectui REPO" knownGaps in studio-authoring.json, and RUNNER's objectui-ref standing fact. Re-scan of the five files' non-history text for objectui:(build|refresh|bump|clean), build-console.sh, .objectui-sha, objectui checkout and "PINNED IN THE objectui REPO": 0 hits. History rows are records and are left as written.

Acceptance notes

• Same stale knownGap in other files, out of scope here. 15 items in records-forms.json (12), automation.json (2) and attachments-storage.json (1) still say their objectui pins are not runnable from this checkout. RUNNER's rewritten fact now tells a runner to read that wording as stale. The items themselves belong to their files' cards, and the report on qa(checklist): i18n — 9 ledger edits from the 18.2.0 release sweep (#25807) #25842 names this for the seat.
• packages/qa/dogfood/test/search-conformance.ledger.ts names showcase-search.dogfood.test.ts as the proof for search-fields-override (400 INVALID_FIELD), but that file has no refusal case. Noted, not filed.
• examples/app-showcase/src/system/self-url.ts's header still says the os dev value follows the live environment; at 6679644a the served metadata did not. This goes to the seat with the OS_PORT finding.

Verification

At head bc394208:

• pnpm check:platform-checklist: exit 0, "OK — 16 areas, 329 items".
• node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derives 13 families. All 13 ran with exit 0: check:doc-formula-expressions refused with exit 3 until @objectstack/formula and @objectstack/lint were built, then passed.
• --ran reconciliation: 13 derived, 13 run, 0 NOT-MEASURED.
• node scripts/check-commit-card-trailers.mjs --range origin/main..HEAD: 5 commits, no card relation, no model identifier.

[objectstack-fleet]

席位核对 · ACCEPT · devx seat 2 · 2026-10-26T12:46Z

Seat domain:devx#2 (#20163) · session_01Q4MDXRbEJCNdPAP4SiXwp9 · reviewed head bc39420839 · family dispatch, chain head #25842 (claims 6415922069, 6415936879, 6415951083, 6415957906, 6415989105, all naming this branch).

• Shape: draft → main; four Fixes lines (qa(checklist): i18n — 9 ledger edits from the 18.2.0 release sweep (#25807) #25842, qa(checklist): search — 6 ledger edits from the 18.2.0 release sweep (#25807) #25847, qa(checklist): studio-authoring — 9 ledger edits from the 18.2.0 release sweep (#25807) #25848, qa(checklist): RUNNER.md — 3 stale environment facts from the 18.2.0 release sweep (#25807) #25850) and, after the seat's body edit, Part of #25849: one of qa(checklist): ux-conventions — 5 ledger edits from the 18.2.0 release sweep (#25807) #25849's five findings (ux-conventions.write-feedback acceptance[1], whether object-only success strings name what changed) waits on a maintainer ruling, so that card must not close on merge. Clause-②: no, Metadata-protocol: none at line start. One commit per card.
• Premise measured: no commit on the five files on origin/main since filing except docs: build the Console with pnpm console:build, not the retired objectui pin scripts #26052 on RUNNER.md, which the dispatch named; the dev re-read on main cc378e93.
• Scope: exactly the five claimed files (+238/−160), all four JSON files parse at the head. No .changeset, no governed path, no contract face ⇒ skip-changeset right, no second-agent review owed.
• Spot readings on origin/main: showcase-search.dogfood.test.ts:46 asserts names.some(... includes('retail')), as the revised search verify now says. Root package.json has console:build = turbo run build:dist --filter=@objectstack/console; scripts/build-console.sh and .objectui-sha are absent; packages/console/build-dist.sh exists. So the retired-pin rewrites in RUNNER and studio-authoring hold.
• Gates: dev-local 13/13 derived families exit 0, check:platform-checklist OK (report 6416286751). CI on this head was running at review; ready waits for every check green or skipped.
• Accepted deviation: the retired-pin rewrite went past regex hits to the "AUTOMATION IS PINNED IN THE objectui REPO" knownGaps of the same files, so the edited items do not contradict themselves. History rows were left as records.
• Out of scope: the class c (translation item name) and class b (os dev served baseUrl) findings will be filed by this seat through the filing gate. The same knownGap in records-forms, automation and attachments-storage was handed to those files' in-flight cards (qa(checklist): records-forms — 60 ledger edits from the 18.2.0 release sweep (#25807) #25846, qa(checklist): automation — 24 ledger edits from the 18.2.0 release sweep (#25807) #25840, qa(checklist): attachments-storage — 18 ledger edits from the 18.2.0 release sweep (#25807) #25839).
• Landing to-do: after MERGED, qa(checklist): ux-conventions — 5 ledger edits from the 18.2.0 release sweep (#25807) #25849 leaves pm:dispatched for needs-user-decision with the A1 question; the other four cards close by Fixes, and their pm:dispatched label comes off in the same act.

---

Generated by Claude Code