docs(qa): 18.0 sweep anchor round 4 — §3 accuracy corrections in studio-authoring, records-forms, dashboards and search by objectstack-fleet[bot] · Pull Request #24671 · objectstack-ai/objectstack

objectstack-fleet · 2026-10-19T22:21:58Z

Refs #24579
Clause-②: no

The fourth dispatch of the 18.0 release sweep's wave anchor. It covers §3's checklist-accuracy findings for the four UI-facing areas: studio-authoring, records-forms, dashboards and search. It takes round 2's additions for them (comments 6272032736 and 6272314386), the Tier 1 automated.ref notes that land in these files, and the leftovers round 3 named.

This PR edits the ledger only. The diff is the four docs/qa/platform-checklist/areas/ files the claim names, with no product code, test or pin in either repository. Each edited item takes one revision bump and one history row that names the run records it rests on. Several findings on one item make one revision. Every finding was re-derived from source before it was edited:

the server at objectstack base f89d29e382, re-checked after merging origin/main bfc1539fe0; no cited source moved in between;
the console at the shipped pin .objectui-sha 4016370af3c9, cross-checked at objectui main 282b138c07, where every cited fact is the same;
the run's own captured evidence, where the record has it.

Findings that did not hold are left alone, and each is listed below with the reading that refutes it.

The history rows, and the recipe's provenance line, carry the placeholder ref #PENDING-24579R4, because the PR has no number until it is created. A follow-up commit on this branch swaps exactly those strings for this PR's number.

Edited items

studio-authoring (11 items)

item	fields edited	revision	run record	finding rests on
`studio-authoring.first-run-loop`	`fixtures.knownGaps[0]`, `steps[1]`, `steps[2]`, `steps[3]`, `steps[6]`	3 → 4	#24571, #24556	objectui 4016370af3c9 (pin; same at main 282b138c07)
`studio-authoring.object-designer-roundtrip`	`steps[4]`, `steps[5]`	4 → 5	#24604	objectui 4016370af3c9 (pin; same at main 282b138c07)
`studio-authoring.view-authoring-live`	`personas[1]`, `fixtures.requires[0]`, `steps[3]`	2 → 3	#24571	objectstack `f89d29e`
`studio-authoring.record-page-roundtrip`	`personas[1]`, `fixtures.requires[0]`, `acceptance[0].verify`, `acceptance[3].verify`, `steps[0]`, `steps[2]`, `steps[3]`, `steps[6]`	4 → 5	#24571, #24604 (§3 bullet)	objectstack `f89d29e`; objectui 4016370af3c9 (pin; same at main 282b138c07)
`studio-authoring.draft-publish-lifecycle`	`acceptance[6].clause`, `acceptance[6].verify`, `steps[2]`	6 → 7	#24604 (6272314386)	objectui 4016370af3c9 (pin; same at main 282b138c07); objectstack `f89d29e`
`studio-authoring.authoring-validation-not-persisted`	`acceptance[1].clause`, `acceptance[1].verify`, `acceptance[3].verify`, `steps[1]`, `steps[3]`	3 → 4	#24571	objectstack `f89d29e`; objectui 4016370af3c9 (pin; same at main 282b138c07)
`studio-authoring.permission-matrix-editor-ux`	`fixtures.requires[1]`, `acceptance[4].evidence`, `acceptance[4].verify`	3 → 4	#24604	objectui 4016370af3c9 (pin; same at main 282b138c07)
`studio-authoring.metadata-diagnostics-sweep`	`fixtures.knownGaps[0]`, `acceptance[6].clause`, `acceptance[6].verify`, `steps[6]`, `steps[8]`	4 → 5	#24604 (6272314386)	run evidence #24604; objectstack `f89d29e`
`studio-authoring.doc-on-app-menu`	`steps[0]`	3 → 4	#24571	the item's own knownGaps[2]
`studio-authoring.multi-package-item-rows`	`steps[2]`, `steps[4]`	1 → 2	#24558	objectstack `f89d29e`
`studio-authoring.save-door-runs-authoring-rules`	`acceptance[1].clause`, `acceptance[1].verify`	1 → 2	#24558 (§3 since:v18 bullet)	objectstack `f89d29e`

records-forms (19 items, inside the 20-item cap)

item	fields edited	revision	run record	finding rests on
`records-forms.crud-roundtrip`	`negative[2]`, `fixtures.provisioning.why`, `acceptance[0].verify`, `acceptance[6].verify`, `steps[2]`	10 → 11	#24572, #24556	objectstack `f89d29e`
`records-forms.field-type-matrix`	`fixtures.knownGaps[2]`, `acceptance[1].evidence`, `acceptance[1].clause`, `steps[2]`	3 → 4	#24616 (6272314386)	objectstack `f89d29e`; objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.list-view-capabilities`	`negative[0]`, `negative[1]`, `steps[4]`, `automated.ref`	3 → 4	#24572, #24616 (6272314386)	objectstack `f89d29e`; objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.form-view-gallery`	`fixtures.requires[0]`, `steps[7]`	3 → 4	#24572	run evidence #24572; objectstack `f89d29e`
`records-forms.conditional-rules-grid`	`fixtures.requires[1]`, `steps[2]`	4 → 5	#24616	objectstack `f89d29e`
`records-forms.related-list-server-pagination`	`negative[0]`, `fixtures.knownGaps[0]`, `acceptance[0].clause`, `acceptance[0].verify`, `steps[1]`, `steps[2]`, `steps[6]`	4 → 5	#24616 (6272314386)	objectstack `f89d29e`; objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.action-location-matrix`	`steps[3]`	11 → 12	#24616	objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.validation-rule-type-matrix`	`fixtures.requires[0]`, `variants[0]`	3 → 4	#24572	objectstack `f89d29e`
`records-forms.gantt-interactions`	`steps[4]`	4 → 5	#24616	run evidence #24616
`records-forms.adhoc-filter-sort-builder`	`negative[1]`, `acceptance[0].clause`, `acceptance[0].verify`, `acceptance[1].clause`, `acceptance[1].verify`, `steps[1]`, `steps[2]`, `steps[7]`	3 → 4	#24616 (6272314386)	objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.calendar-interactions`	`acceptance[3].clause`, `acceptance[3].verify`, `steps[3]`, `steps[5]`	2 → 3	#24616 (6272314386)	objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.record-discussion-mentions`	`acceptance[3].verify`, `acceptance[4].clause`, `acceptance[4].verify`, `steps[3]`, `steps[6]`, `source[2]`	3 → 4	#24616 (6272314386)	objectstack `f89d29e`; objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.grid-personalization`	`acceptance[1].clause`, `acceptance[1].verify`, `steps[1]`	4 → 5	#24616 (6272314386)	objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.bulk-select-all-matching`	`acceptance[2].clause`, `steps[5]`	3 → 4	#24616	objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.import-job-undo-cancel`	`fixtures.knownGaps[2]`, `traps[2]`, `steps[1]`	4 → 5	#24616 (6272314386)	objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.object-hook-lifecycle`	`negative[2]`, `fixtures.requires[3]`, `acceptance[8].evidence`, `acceptance[8].verify`, `steps[0]`, `automated.ref`	5 → 6	6272032736, #24598, #24616	objectstack `f89d29e` (4 pins measured)
`records-forms.field-history-tracking`	`negative[0]`, `acceptance[2].verify`	1 → 2	#24616 (6272314386)	objectui 4016370af3c9 (pin; same at main 282b138c07)
`records-forms.comment-reactions-and-edit-stamps`	`fixtures.requires[2]`	1 → 2	#24558	run evidence #24558
`records-forms.v18-semantic-defaults`	`personas[1]`, `fixtures.knownGaps[0]`, `acceptance[2].clause`, `steps[0]`, `steps[1]`	1 → 2	#24558, #24616 (6272314386)	objectstack `f89d29e`; objectui 4016370af3c9 (pin; same at main 282b138c07)

dashboards (5 items)

item	fields edited	revision	run record	finding rests on
`dashboards.strict-widget-rejects-stray-keys`	`steps[4]`	2 → 3	#24573	objectui 4016370af3c9 (pin; same at main 282b138c07)
`dashboards.chart-type-matrix`	`fixtures.knownGaps[0]`, `acceptance[2].clause`, `acceptance[4].verify`, `steps[5]`	5 → 6	#24573 (§3 bullet)	objectstack `f89d29e`; objectui 4016370af3c9 (pin; same at main 282b138c07)
`dashboards.dataset-report-authoring`	`negative[1]`	2 → 3	#24612 (6272314386)	objectstack `f89d29e`
`dashboards.global-filters-rescope`	`fixtures.requires[2]`	4 → 5	#24573	objectui 4016370af3c9 (pin; same at main 282b138c07)
`dashboards.analytics-exposure-gate-and-measure-aggregate`	`acceptance[1].clause`, `acceptance[3].verify`, `steps[0]`, `steps[3]`	1 → 2	#24558	objectstack `f89d29e`

search (3 items and the qa-contributor-bound-member recipe)

item	fields edited	revision	run record	finding rests on
`search.pinyin-flag-both-sides`	`fixtures.requires[3]`, `fixtures.knownGaps[0]`, `acceptance[3].evidence`, `acceptance[3].verify`, `steps[6]`	4 → 5	#24574	objectstack `f89d29e`
`search.console-global-search`	`fixtures.requires[0]`, `acceptance[0].clause`, `acceptance[2].clause`, `steps[2]`	4 → 5	#24574 (§3 bullet)	objectui 4016370af3c9 (pin; same at main 282b138c07)
`search.global-search-skips-unreadable`	`fixtures.requires[2]`, `steps[4]`, `steps[5]`	2 → 3	#24593 (6272314386)	objectstack `f89d29e`
recipe `search:qa-contributor-bound-member`	`fixtures.qa-contributor-bound-member.provenance`, `fixtures.qa-contributor-bound-member.requires[1]`, `fixtures.qa-contributor-bound-member.sequence[1]`, `fixtures.qa-contributor-bound-member.sequence[2]`, `fixtures.qa-contributor-bound-member.sequence[4]`	n/a (recipes carry no revision)	#24574, #24593	objectstack `f89d29e`
The verbatim before and after of every field is this PR's diff, `git diff bfc1539fe0 HEAD -- docs/qa/platform-checklist/areas/`. Each item's history row says what changed and why.

Findings left alone, each with the reading that refutes it

studio-authoring

draft-publish-lifecycle, the draft-discard note ("discarding a draft uses ?state=draft"): true of the door, since META_ITEM_RESET_PARAMS is state, dropStorage, package and member. But no step of any studio-authoring item discards a draft. The only DELETEs in the file are overlay resets, so the note has no clause to correct.
draft-publish-lifecycle, "a scratch dashboard in a namespaced package needs the NS_ prefix": refuted. The prefix rule judges objects only. packages/spec/src/kernel/namespace-prefix.ts says "every object.name", and publish-drafts' preflight skips non-object drafts (if (d.type !== 'object') continue).
draft-publish-lifecycle, rollback refuses ?package=: true, but no step sends it, so there is no item claim to correct.
expression-editors, both QA run · studio-authoring (15/20) · ba4afcc2 · 2026-10-19 · 13 PASS / 1 PARTIAL / 1 FAIL / 0 BLOCKED / 5 NOT-RUN #24604 notes (validate_expression has no RLS site; the package door has no Save button): already on main through docs(qa): 18.0 sweep anchor round 2 — expression-editors' RLS clauses to the engine's error tier, a consumed-changeset citation, and §3 accuracy corrections in five areas #24632 (revision 4).
first-run-loop, step 6's Create-app checkbox already adds the nav item: already stated by the item's knownGaps[1] (SUPERSEDED, QA run QA run · priority:P0 (19/19) · ef1ed17f · 2026-10-01 · 13 PASS / 0 PARTIAL / 6 FAIL / 0 BLOCKED / 0 NOT-RUN #21056).
record-page-roundtrip, "a record:highlights block without fields passes the draft PUT but fails publish": an observation with no clause behind it. The item adds a Card, not an empty highlights block.
permission-matrix-editor-ux, the filter "owner" matches only owner_id: no clause names a filter term.
packaged-automation-studio-lock, the disabled switch looks green, and canvas nodes are not React Flow nodes: cosmetic, plus a locator hint. The item names neither a colour nor React Flow (0 hits).
object-designer-roundtrip, live spec F4/F1 stale: §2's objectui-live-e2e theme (see below).
doc-on-app-menu and authoring-validation-not-persisted: the objectui observations (empty main area, menu highlight, the "Auto-saving…" pill) are product observations, not item claims.

records-forms

crud-roundtrip, the Tier 1 note that its automated.ref over-claims: refuted. The ref already scopes itself: "acceptance[6]'s READONLY half ONLY … Every other clause still needs the run". QA run · tier1:automated-pins (149/166) · ba4afcc2 · 2026-10-19 · 60 PASS / 89 PARTIAL / 0 FAIL / 0 BLOCKED / 17 NOT-RUN #24598 read the same: only the readonly half of clause 6.
object-hook-lifecycle, "the steps' $select — the by-id read door accepts select only": no step of this item spells $select, and records-forms.json has 0 hits.
list-view-capabilities negative[0], "the refusal code is not INVALID_FILTER": refuted at source. packages/objectql/src/boolean-comparand-declared-type-door.ts refuses any non-boolean string against a boolean field with INVALID_FILTER / 400, naming the field, and that file is unchanged since ba4afcc2. The finding's staging recipe was taken.
list-view-capabilities step 7, the switcher lives under Display → Layout: a locator hint. The step names no location.
validation-rule-type-matrix, QA run · since:v18 (48/48) · 5b14b7f3 · 2026-10-19 · 38 PASS / 7 PARTIAL / 3 FAIL / 0 BLOCKED / 0 NOT-RUN #24558's "describes the advisory hit as server-log only": already on main (revision 3 reads the write answer's warnings).
field-group-visible-when, the item cites the console pin 1fb1ced36243, now 4016370af3c9: a dated reading that still holds at the shipped pin. plugin-form/src/fieldGroups.ts passes visibleWhen through, buildDefaultPageSchema.ts writes { group } references, and both groupVisibleWhen-11630 pins exist. The record's other two notes (the grouped section renders above the ungrouped fields; an invalid CEL visibleWhen is accepted at authoring) are observations. No clause claims either one.
predicate-relationship-traversal, "use filter= JSON, not an OData $filter=": no step of this item sends a list filter.
cascading-options, the invite recipe needs accept-invitation: already stated by the item's personas[1] ("a member invited with role 'admin' who has ACCEPTED the invitation"). The RUNNER.md recipe is outside the file surface (see below).
import-wizard-encoding-and-hints step 3, Contacts' Import entry sits in the overflow menu: no step names the entry's location.
concurrent-edit-conflict, the dialog buttons read "Reload latest / Overwrite anyway / Cancel": the item names the two actions by their verbs (Reload / Overwrite), which is accurate shorthand. It is also one of the seven hand-driven now-PASS items, and its stated clauses hold.
cascading-multilevel-and-clear (test ids), lookup-picker-create-new (the typed term is not pre-filled, an observation by the record's own words), bulk-select-all-matching (the header search opens the palette), adhoc-filter-sort-builder ("Add filter" persists an incomplete row) and gantt-interactions (null-date tasks drawn as milestones): locator hints or product observations, with no item claim behind them.
view-type-gallery, the map base style is unreachable from the sandbox: an environment fact, not an item claim. The RUNNER environment table is outside the file surface.
action-param-widgets, the p_richtext helpText contradicts the textarea widget: that copy is showcase metadata, which is product code, and the ruling forbids product edits.
record-discussion-mentions, the mentioned member sees "You don't have permission to view comments": a possible by-design observation, with no clause. acceptance[8]'s scratch-object channel is already in steps[12].
v18-semantic-defaults, "the canonical file-DB boot works too": a suggestion.

dashboards

strict-widget-rejects-stray-keys, "a pending draft is unreadable until publish": refuted. GET /meta/dashboard/NAME?state=draft reads the draft (META_ITEM_READ_PARAMS includes state). Only ?mode=draft, a save-door parameter, is refused on the read. The inspector half was taken.
empty-null-bucket-boundaries, the 79-click 2020 window, the two-month seed and the single-bucket recipe: runner-efficiency suggestions. The step's "a week in 2020" works, and the seed-data-thin and single-datapoint traps are already listed.
chart-type-matrix, a thin seed for trend charts: the single-datapoint trap is already listed.
global-filters-rescope, the inspector labels the date binding dateRange: cosmetic.
analytics-exposure-gate-and-measure-aggregate, "stock has no cube over an unexposed object — scratch app required": already stated by fixtures.requires[1]–[2].
The console route spelling note: both spellings work, and no item claims otherwise.

search

pinyin-flag-both-sides, the lookup-picker part of step 5 was not exercised: run coverage, not an inaccuracy.
§3's "notify assignee must be a user id; the job timeout probe needs a retryPolicy": these are integration-system.notify-inbox-delivery and integration-system.job-scheduled-run (QA run QA run · integration-system (16/20) · 5b14b7f3 · 2026-10-19 · 11 PASS / 5 PARTIAL / 0 FAIL / 0 BLOCKED / 4 NOT-RUN #24575), and integration-system.json is excluded (open PR feat(plugin-webhooks,runtime)!: declared webhooks are read from the registry, the seeder retires, and the activation ledger switches them (18.x step 1, 8b-0g′) #24588).

Named, not acted on

§2 objectui-live-e2e theme: these items cite red or stale objectui live specs in automated.ref and were left there: studio-authoring.object-designer-roundtrip (studio-object-designer F1/F4), studio-authoring.record-page-roundtrip (already marked automated.stale), records-forms.list-view-capabilities (user-filters, inline-edit-polish-2572; only its framework dogfood half was qualified), records-forms.adhoc-filter-sort-builder (user-filters), records-forms.conditional-rules-header (field-conditional-rules), records-forms.conditional-rules-grid (grid-file-upload), records-forms.cascading-options and records-forms.cascading-multilevel-and-clear (cascading-options.spec:51), records-forms.master-detail-atomic-save (form-view-subforms), records-forms.related-list-server-pagination (detail-related-list), records-forms.concurrent-edit-conflict (inline-edit-polish-2572), records-forms.field-history-tracking (record-history-display, which always skips against showcase).
§2, pins that cover no full clause: records-forms.action-param-widgets' objectui half, list-row-action-cel.spec.ts for action-location-matrix / conditional-rules-header, and the seven hand-driven now-PASS items. Their "pin covers no clause" side is left alone. The stated-clause errors on five of them (related-list, adhoc, calendar, grid-personalization, import-job) were taken.
§4 fixture gaps: records-forms.validation-rule-type-matrix negative[0] (no door authors a non-evaluable CEL rule).
Outside the file surface: RUNNER.md's "Invite, then sign up" recipe invites with role member and does not say that a non-member role takes effect only after POST /api/v1/auth/organization/accept-invitation (QA run QA run · records-forms (38/44) · ba4afcc2 · 2026-10-19 · 34 PASS / 2 PARTIAL / 2 FAIL / 0 BLOCKED / 6 NOT-RUN #24616). RUNNER's environment table would also take the sandbox map-tile fact.
Over the cap: none. records-forms took 19 of its 20.

Measurements

PM hypothesis 3, the object-hook-lifecycle ref: it named source files, hook-wrappers.ts and hook-binder.ts, which no resolver can run. It now names four test files. Each was run at base f89d29e382 under scripts/pm/os-verify-lock.sh (VERDICT command-exit 0), after building @objectstack/objectql's dependency closure under the same lock (13/13 tasks):

file tests

hook-condition-previous-scope.test.ts 25

hook-condition-merged-record.test.ts 12

hook-binder.test.ts 26

hook-condition-fail-loud.test.ts 20

That is 83 passed, 0 failed. Each file is mapped in the ref to the clause whose mechanism it asserts: acceptance[3], [5], [7] and [8].
PM hypothesis 1: Studio RLS editor rates a non-lowerable USING predicate a WARNING ("can … WIDEN access") and autosaves it, while publish / os build refuse it as an error (rls-predicate-unenforceable, fail-closed) objectui#13061, [finding] os serve publishes the REQUESTED port, not the bound one — with --port 0 it announces a port nothing is listening on, and the code comment asserts the opposite #13062 and Three broken-sweep mentions in packages/spec/src/automation still call the predicate an alert / a query that can tell — the residue outside the fences of #12722 and #12900 #13063 are in state completed. Their repairs are at objectui main, after the shipped pin 4016370af3c9, so none of them makes a clause true again at the pin, and their FAIL clauses keep their product-defect disposition. 18.0 round-2 sweep framework findings: a paid invoice line's amount/total stays writable while its inputs are locked; an unevaluable hook condition answers 500 INTERNAL_ERROR; remote-tables?schema= never narrows on the SQL drivers #24619 is open. Its row (b) is why object-hook-lifecycle acceptance[8] now states the shipped 500 envelope and scores against 18.0 round-2 sweep framework findings: a paid invoice line's amount/total stays writable while its inputs are locked; an unevaluable hook condition answers 500 INTERNAL_ERROR; remote-tables?schema= never narrows on the SQL drivers #24619 rather than re-filing it.
PM hypothesis 2: no disagreement between the pin and objectui main for any cited console fact. All 25 cited objectui files were diffed between 4016370af3c9 and 282b138c07, and the facts in the changed ones were re-read at main.
PM hypothesis 4: check:platform-checklist symbol anchors stay at 875/885, the count at merged origin/main bfc1539fe0 after docs(qa): 18.0 sweep anchor round 3 — the approvals token door after the reminder link removal, a consumed-changeset citation, and §3 accuracy corrections in six areas #24649 landed: 0 lost, 10 on the [finding] the platform-checklist corpus resolves symbol anchors with its OWN rule, not the shared resolver — a permissive token match where the ruling says there is to be exactly one implementation #16898 residual, 18 file floors held.
Gates: the list was derived with node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack, which gave 13 families for 4 paths against merge base bfc1539fe. All 13 exited 0 at cccfc9fcb2, each code captured before any pipe. --ran reports: 13 derived, 13 run, 0 NOT-MEASURED. check:platform-checklist runs in no per-PR CI, so this local reading is its evidence: check-platform-checklist: OK — 16 areas, 327 items (322 active, 1 planned).
Every area-file edit went through a span-recording textual editor. It re-parses the file and asserts three things: the document equals the old one with the declared operations applied, every other item is unchanged in value, and each edited item's changed leaf paths stay within the declared fields, the revision and one new history row.
skip-changeset, re-measured: docs/qa/** is in no package's files[], so this diff publishes nothing.

Acceptance notes

studio-authoring.record-page-roundtrip moves its object from showcase_invoice to showcase_contact (§3's own prescription). Round 1's acceptance[0]–[2] verdicts were read on showcase_invoice at revision 4, and revision 5 scores them on showcase_contact.
records-forms.crud-roundtrip's non-owner clone recipe (a contributor member cloning a seeded showcase_field_zoo row) is source-derived and not yet run. The item says so and asks the run to record the substitution.
Observed, not filed: the record notes that an invalid CEL visibleWhen on fieldGroups is accepted at authoring with 200 (QA run QA run · records-forms (30/44) · 5b14b7f3 · 2026-10-19 · 11 PASS / 19 PARTIAL / 0 FAIL / 0 BLOCKED / 14 NOT-RUN #24572). The PM may want to route it.

Generated by Claude Code

…accelerator toggle, the pinyin roster oracle, global search's perObject cap and stock hidden value, and the contributor recipe's admin doors Checklist ledger only. Three search items take one revision each and the qa-contributor-bound-member recipe names the REST doors QA run #24574 measured; every finding re-read at objectstack f89d29e and, for the console, at the shipped pin 4016370af3c9. Claude-Session: https://claude.ai/code/session_01FM29ZqLTrnPNXxijioZwMp Co-authored-by: Claude <noreply@anthropic.com>

…o free-key door, combo marks are the renderer's, the Interfaces pillar needs an app, the dangling-binding guard sits at publish, and the unexposed cube answers CUBE_NOT_FOUND Checklist ledger only. Five dashboards items take one revision each, every finding re-read at objectstack f89d29e and the console pin 4016370af3c9. Claude-Session: https://claude.ai/code/session_01FM29ZqLTrnPNXxijioZwMp Co-authored-by: Claude <noreply@anthropic.com>

…re-pointed to the shipped console and server (review sheet, autosave, record page on showcase_contact, draft line, Source-tab door, package-scoped matrix, plantable diagnostics row, owner-mismatch and publish-drafts doors, rule ids) Checklist ledger only. Eleven studio-authoring items take one revision each, every finding re-read at objectstack f89d29e and the console pin 4016370af3c9; findings that did not hold are left alone. Claude-Session: https://claude.ai/code/session_01FM29ZqLTrnPNXxijioZwMp Co-authored-by: Claude <noreply@anthropic.com>

…e-pointed to the shipped console and server (wire params without a $, three densities, the server-backed mention bell, hook route and refs, the shipped 500 for an unevaluable hook condition, autonumber off the form, the crypto provider on every CLI boot, and run-proven fixture recipes) Checklist ledger only. Nineteen records-forms items take one revision each, inside the dispatch's 20-item cap; every finding re-read at objectstack f89d29e and the console pin 4016370af3c9. The object-hook ref now names four test files measured at base (83 passed). Claude-Session: https://claude.ai/code/session_01FM29ZqLTrnPNXxijioZwMp Co-authored-by: Claude <noreply@anthropic.com>

…8-sweep-accuracy-round4

Swaps the 39 '#PENDING-24579R4' placeholders (38 history rows and the qa-contributor-bound-member recipe's provenance line) for #24671. Nothing else changes. Claude-Session: https://claude.ai/code/session_01FM29ZqLTrnPNXxijioZwMp Co-authored-by: Claude <noreply@anthropic.com>

objectstack-fleet · 2026-10-19T22:34:37Z

Seat review

ACCEPT · domain:devx seat 1 · os-warren · session_01FM29ZqLTrnPNXxijioZwMp · 2026-10-19T22:34Z · head 7d32eecf3b · card #24579 round 4 (claim 6277591296)

Form. Draft, base main. First lines are Refs #24579 and Clause-②: no. Assignee is os-warren; labels are size/l and skip-changeset. node scripts/pm/check-governed-merges.mjs --pr 24671 → NOT governed, 553 changed lines (≤ 3000). git merge-tree against main bfc1539fe0 is clean.
Scope. The four claimed files under docs/qa/platform-checklist/areas/, +372 / −181, and no other path. No package, test, pin or gate is touched, in either repository.
Ledger invariants, checked by parse against the merge base bfc1539fe0.
- 38 items differ, and each took exactly one revision bump and one appended history row with ref docs(qa): 18.0 sweep anchor round 4 — §3 accuracy corrections in studio-authoring, records-forms, dashboards and search #24671 (studio-authoring 11, records-forms 19, dashboards 5, search 3).
- No item was added or removed.
- The one area-level change is search.json's qa-contributor-bound-member recipe. A recipe carries no revision, so its provenance line records the change (round 2's precedent).
Spot-checked.
- records-forms.object-hook-lifecycle's ref now names four existing tests in place of two source files: hook-condition-previous-scope, hook-condition-merged-record, hook-binder and hook-condition-fail-loud, with 83 passing under the verify lock. Each is mapped to the clause it asserts.
- The density note reads three modes, and the shipped console pin 4016370af3c9's useDensityMode names compact, comfortable and spacious.
- The report names, for every edited item, the repo and sha its finding rests on. Console facts were read at the shipped pin and cross-checked at objectui main, and none differs (PM H2 falsified).
Gates.
- The dev's dispatch-gates union: 13 derived, 13 run, all exit 0 at 7d32eecf3b.
- check:platform-checklist holds 875/885 anchors, none lost.
- Check-runs at 7d32eecf3b: 39, all completed, 31 success and 8 skipped. node scripts/pm/check-expected-skips.mjs --pr 24671 → OK — 8 skipped check-run(s), every one in the roster (exit 0).
- No review and no comment.
Left alone, accepted. The report lists each finding it did not take, with the refuting reading. Two examples:
- list-view-capabilities negative[0]'s "not INVALID_FILTER" is refuted by boolean-comparand-declared-type-door.ts.
- strict-widget-rejects-stray-keys's "draft unreadable until publish" is refuted by META_ITEM_READ_PARAMS carrying state.
Carried on the card, for the next round.
- RUNNER.md's "Invite, then sign up" recipe, which omits accept-invitation for a non-member role.
- The §6 environment fact on unreachable map tiles.
- The integration-system notify/job notes, which wait on PR feat(plugin-webhooks,runtime)!: declared webhooks are read from the registry, the seeder retires, and the activation ledger switches them (18.x step 1, 8b-0g′) #24588.
- The §2 objectui-live-e2e refs the report names.

This seat readies the PR and queues it with auto-merge.

Generated by Claude Code

claude added 5 commits October 19, 2026 22:03

Merge remote-tracking branch 'origin/main' into claude/issue-24579-v1…

cccfc9f

…8-sweep-accuracy-round4

objectstack-fleet Bot assigned os-warren Oct 19, 2026

objectstack-fleet Bot added the skip-changeset PR has no user-facing published change; bypasses the changeset gate label Oct 19, 2026

github-actions Bot added the size/l label Oct 19, 2026

objectstack-fleet Bot mentioned this pull request Oct 19, 2026

qa(checklist): 18.0 release sweep anchor at 5b14b7f3 — re-point 5 red-row clauses + the wave's checklist-accuracy findings, fixture gaps and not-run handover (18 records) #24579

Open

objectstack-fleet Bot marked this pull request as ready for review October 19, 2026 22:35

objectstack-fleet Bot enabled auto-merge October 19, 2026 22:35

objectstack-fleet Bot added this pull request to the merge queue Oct 19, 2026

Merged via the queue into main with commit 3a931fc Oct 19, 2026
40 checks passed

objectstack-fleet Bot deleted the claude/issue-24579-v18-sweep-accuracy-round4 branch October 19, 2026 23:01

github-actions Bot mentioned this pull request Oct 20, 2026

hourly full run: red on main (CI) #24700

Closed

objectstack-fleet Bot mentioned this pull request Oct 20, 2026

[PM seat] domain:devx @ objectstack — 🟢 os-warren · session_01FM29ZqLTrnPNXxijioZwMp #6023

Open

Navigation Menu

Repository navigation

docs(qa): 18.0 sweep anchor round 4 — §3 accuracy corrections in studio-authoring, records-forms, dashboards and search - #24671

docs(qa): 18.0 sweep anchor round 4 — §3 accuracy corrections in studio-authoring, records-forms, dashboards and search#24671
objectstack-fleet[bot] merged 6 commits into
mainfrom
claude/issue-24579-v18-sweep-accuracy-round4

objectstack-fleet Bot commented Oct 19, 2026

Uh oh!

objectstack-fleet Bot commented Oct 19, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

file	tests
`hook-condition-previous-scope.test.ts`	25
`hook-condition-merged-record.test.ts`	12
`hook-binder.test.ts`	26
`hook-condition-fail-loud.test.ts`	20

Conversation

objectstack-fleet Bot commented Oct 19, 2026

Edited items

studio-authoring (11 items)

records-forms (19 items, inside the 20-item cap)

dashboards (5 items)

search (3 items and the qa-contributor-bound-member recipe)

Findings left alone, each with the reading that refutes it

Named, not acted on

Measurements

Acceptance notes

Uh oh!

objectstack-fleet Bot commented Oct 19, 2026

Seat review

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants