QA: close out the 14 open qa-run records of the e4e5c6e3 wave — supersede sweep, four-way FAIL disposition, extraction

[os-zhuang]

Maintainer authorization (2026-08-18, PM chat, skills-seat session session_01Rn7aaamsR99FXRqLcpL99q, direct-dispatch channel): 「同意你的落地动作」, approving the QA-loop landing plan whose item 1 is this card: process the existing stock of open qa-run records.

Coordination note: these records belong to the #9296 wave (pm:epic, owning session named on that card). This close-out executes under the maintainer authorization above, announced by comment on #9296 — it implements that card's own standing rule ("defects extract into their own cards, which is the PM's close-out duty"), it does not re-run or re-judge any round.

Scope

The 14 open qa-run issues as of 2026-08-18: #9330 #9332 #9333 #9334 #9336 #9337 #9338 #9351 #9352 #9353 #9401 #9417 #9453 #9467.

Procedure per record

1. Supersede check first (abandoned R3 area runs: QA run · access-security (9/20 items consulted) · e4e5c6e3 · 2026-08-17 · 1 PASS / 7 PARTIAL / 0 FAIL / 2 BLOCKED / 10 NOT-RUN #9330 QA run · identity-auth (4/14 items consulted) · e4e5c6e3 · 2026-08-17 · 0 PASS / 4 PARTIAL / 0 FAIL / 2 BLOCKED / 8 NOT-RUN #9332 QA run · api-backend (8 of 14 items) · e4e5c6e3 · 2026-08-17 · 4 PASS / 4 PARTIAL / 0 FAIL #9333 QA run · automation (5 of 10 items) · e4e5c6e3 · 2026-08-17 · 1 PASS / 3 PARTIAL / 1 BLOCKED #9336 QA run · platform-core (4/13 items consulted) · e4e5c6e3 · 2026-08-17 · 1 PASS / 3 PARTIAL / 0 FAIL / 9 NOT-RUN #9337 QA run · integration-system (6/14 items consulted) · e4e5c6e3 · 2026-08-17 · 2 PASS / 4 PARTIAL / 0 FAIL / 8 NOT-RUN #9338 QA run · records-forms (FULL area) · e4e5c6e3 · 2026-08-17 · 2 PASS / 10 PARTIAL / 1 FAIL #9351 QA run · approvals (FULL area) · e4e5c6e3 · 2026-08-17 · 0 PASS / 1 PARTIAL / 0 FAIL (10 not-run) #9352 QA run · studio-authoring (FULL area) · e4e5c6e3 · 2026-08-17 · 0 PASS / 1 PARTIAL / 0 FAIL (11 not-run) #9353): the tier runs (QA run · tier1:automated-pins (77 items) · e4e5c6e3 · 2026-08-17 · 35 PASS / 39 PARTIAL / 0 FAIL #9401 tier1-pinned, QA run · tier2a:api-cli-build (23/23) · e4e5c6e3 · 2026-08-17 · 15 PASS / 6 PARTIAL / 1 FAIL #9417/QA run · tier2b:browser-1 (21/21) · e4e5c6e3 · 2026-08-18 · 7 PASS / 12 PARTIAL / 2 FAIL #9453/QA run · tier2c:browser-2 (18/18) · e4e5c6e3 · 2026-08-18 · 10 PASS / 4 PARTIAL / 4 FAIL #9467 tier2 P0/P1-unpinned) re-covered most of their ground. Verify overlap per record, do not assume: an area run's consulted items that are P2-unpinned were NOT re-covered by any tier run (Tier-3 was deferred by decision) — any verdict a tier run did not re-derive must be carried forward into the close-out comment, not silently dropped. Then close as not_planned with a Superseded-by: #<tier-run> line naming what carried each verdict forward.
2. Four-way FAIL disposition (completed runs: QA run · priority:P0 · e4e5c6e3 · 2026-08-17 · 7 PASS / 7 PARTIAL / 2 FAIL / 2 BLOCKED #9334 R1-P0, QA run · tier2a:api-cli-build (23/23) · e4e5c6e3 · 2026-08-17 · 15 PASS / 6 PARTIAL / 1 FAIL #9417, QA run · tier2b:browser-1 (21/21) · e4e5c6e3 · 2026-08-18 · 7 PASS / 12 PARTIAL / 2 FAIL #9453, QA run · tier2c:browser-2 (18/18) · e4e5c6e3 · 2026-08-18 · 10 PASS / 4 PARTIAL / 4 FAIL #9467, and any non-superseded FAIL from step 1): classify each FAIL on the record's own evidence — (a) product defect → file an independent bug card (dup-check first; fix-landing repo per file-at-destination — console/UI defects go to objectstack-ai/objectui); (b) assertion/spec defect → file a test-fix card in the repo owning the spec; (c) stale spec → file a re-point card; (d) invalid pairing / environment → record as blocked in the close-out comment, no card.
3. Extracted cards: filed WITHOUT pm labels and WITHOUT assignee (they enter normal triage first-touch), body carries a greppable QA-source: #<run> · <area.item> · <clause> line. Note QA run · records-forms (FULL area) · e4e5c6e3 · 2026-08-17 · 2 PASS / 10 PARTIAL / 1 FAIL #9351 already extracted 【缺陷】被 multiple:true lookup 指向的对象 REST DELETE 全 400——cascadeDeleteRelations 依赖探针用裸等值查 JSON 列 #9362 — link, don't duplicate.
4. PARTIAL gaps: do NOT file cards (three security-priority cards are already filed by the PM; the rest are parked on tracking: full platform-checklist regression against main @ e4e5c6e3 — five-round orchestration #9296 by the PM). Reference the parking comment.
5. Security findings: per the maintainer disclosure ruling (recorded on skills: five principle gaps the #9296 wave exposed in checklist-test / checklist-author / dogfood-verification #9387 and in tracking: full platform-checklist regression against main @ e4e5c6e3 — five-round orchestration #9296's header): existence only, never a reproduction, anywhere on GitHub.
6. Close-out comment shape on every record before closing, machine-greppable: Extracted: #a #b … / Carried-forward: <verdicts> / Parked-on: #9296 / Superseded-by: #N (whichever apply), then close (completed for extracted runs, not_planned for superseded ones).

Constraints

• GitHub issue operations only — no repo files, no code, no PRs.
• Every new card dup-checked before filing (keyword + error-string searches).
• Report back: per-record disposition table + list of filed cards.

[os-zhuang]

Claim: PM loop round 3 (skills seat, maintainer-present session, direct-dispatch authorization quoted in the card body)
Session: session_01Rn7aaamsR99FXRqLcpL99q
Branch: none — no file surface; this card is GitHub issue operations only (no code, no PRs, no repo writes)
Worktree: none
Domain: domain:skills (QA-loop process execution under this seat, per the 2026-08-18 authorization)
Container & model: M, issue-ops subagent (general-purpose), model: opus — no file surface ⇒ no path-derived tier mandate (dispatch-gates line quoted in prior claims this round); judgment work is FAIL classification, priced at the default judgment tier
Clause-②: no
Serial constraints cleared: no file surface ⇒ no same-file constraints; does not count against the dev batch's merge-queue load; coordination with the #9296 wave territory recorded on that card

---

Generated by Claude Code

[os-zhuang]

Close-out complete — all 14 open qa-run records processed and closed. label:qa-run is:open now returns 0. Nothing was re-run and no verdict was re-judged; every record carries a close-out comment with the machine-greppable lines before its close.

Per-record disposition

record	scope	action	cards	carried-forward
#9330	access-security, 9/20	superseded → not_planned	—	share-link-capability-tokens = blocked(fixture, ledger), P2 unpinned
#9332	identity-auth, 4/14	superseded → not_planned	—	oauth-app-consent-loop, linked-accounts-social = blocked(fixture, ledger), both P2 unpinned
#9333	api-backend, 8/14	superseded → not_planned	—	date-range-preset-matrix = pass (P2 unpinned); query-contract-matrix C0–C3/C5–C8 live evidence (Tier-1's pin reaches C4 only)
#9334	R1 priority:P0, 18/18	extracted → completed	linked #9362, #9391	—
#9336	automation, 5/10	superseded → not_planned	—	flow-toggle-kill-switch = pass (P2 unpinned)
#9337	platform-core, 4/13	superseded → not_planned	—	none — all four re-derived
#9338	integration-system, 6/14	superseded → not_planned	—	none — all six pinned, all re-derived by Tier-1
#9351	records-forms, 14/32	superseded → not_planned	linked #9362 (extracted 2026-08-17)	named-import-mapping = pass (P2 unpinned); related-list-server-pagination c1 = pass (live; Tier-1 scored the item blocked — pin unconsultable in a matched pairing)
#9352	approvals, 1/11	superseded → not_planned	—	none
#9353	studio-authoring, 1/12	superseded → not_planned	—	none
#9401	Tier-1 pins, 77/77	extracted → completed	linked #9481, #9482, #9483	— (all remaining gaps parked by pointer)
#9417	Tier-2A, 23/23	extracted → completed	filed #9487, #9488; linked #9418	—
#9453	Tier-2B, 21/21	extracted → completed	linked #9454, #9362	—
#9467	Tier-2C, 18/18	extracted → completed	linked #9454, #9468, #9469, #9391	—

Supersede check was done per record, not assumed. Coverage of the tier runs = the 77 pinned items (#9401) ∪ the 62 unpinned P0/P1 items enumerated in #9417 / #9453 / #9467. Every consulted item in the nine abandoned area runs was matched against that set individually. The four items that fall outside it are all P2-unpinned — exactly the Tier-3 hole the card predicted — and are named in the table above; two further verdicts are carried because the tier run's own coverage was narrower than the area run's (query-contract-matrix live clauses, related-list-server-pagination c1).

FAIL disposition — 9 FAIL clauses across the four completed runs, all class (a)

FAIL	class	destination
#9334 · records-forms.crud-roundtrip c2	(a) product	#9362 (open, fix PR #9437)
#9334 · integration-system.datasource-credential-refusal-matrix c7	(a) product, access-control	#9391 — existence + item/clause only
#9417 · cli.migrate-meta-codemod c1+c2	(a) product	#9418
#9453 · records-forms.crud-roundtrip c2	(a) product	#9362 (dup — third derivation)
#9453 · studio-authoring.view-authoring-live c3	(a) product	#9454
#9467 · integration-system.datasource-credential-refusal-matrix c7	(a) product, access-control	#9391 — existence + item/clause only
#9467 · dashboards.strict-widget-rejects-stray-keys c5	(a) product	#9454 (same mechanism)
#9467 · dashboards.global-filters-rescope c6	(a) product	#9454 (same mechanism)
#9467 · i18n.notification-localized-and-clears c1	(a) product	#9468
#9467 · R-4 (fails no clause)	(a) product	#9469

No FAIL classified as (b) assertion/spec defect or (c) stale spec — the wave's spec problems all surfaced as partial/CF findings, not as failures. Class (d) entries (recorded blocked, no card) are written into the close-out comments: the no-active-org-session-semantics tenancy-wall dead end, saved-report-ownership's absent plugin-reports, Tier-1's 11 objectui e2e/live reds under an unmatched pairing plus its 3 blocked items and 1 enterprise-gated skip. No reproduction of an access-control defect was written anywhere.

Issues filed by this close-out — two

• Every 401 answers {error, message} with no code key, while every other error family carries {error, code} #9487 — every 401 answers {"error":"UNAUTHENTICATED","message":…} with no code key, while the other six error families answer {error, code}; a client keying on body.code reads undefined for every auth failure, and it already cost QA run · api-backend (8 of 14 items) · e4e5c6e3 · 2026-08-17 · 4 PASS / 4 PARTIAL / 0 FAIL #9333 two clauses.
• GET /api/v1/meta/<unknown-type> answers 200 with an empty collection while the write door refuses the same type #9488 — GET /api/v1/meta/<unknown-type> answers 200 {"items":[]}, indistinguishable from a real-but-empty type, while the write door for the same name refuses precisely (meta-unknown-type-namespace: a /meta type name that is not a plural of anything still mints a namespace — PUT /meta/fieldz/x answers 200 #8421 fixed only the write side).

Both were designated for extraction by the Tier-2A reviewer of record ("worth their own cards rather than burial in a run record") and were the only defect findings in the wave with no card. Both dup-checked by keyword and error string, both carry a QA-source: line, both filed with bug only — no pm:* label, no assignee, no priority — so they enter normal triage first-touch. Everything else the records name was already carded: #9454, #9468 and #9469 had already been extracted from Tier-2B/2C at review time, so the dup rule applied and nothing was re-filed.

Residuals — for your decision, not done here

1. Three records still carry work labels the wave's own rule forbids: QA run · priority:P0 · e4e5c6e3 · 2026-08-17 · 7 PASS / 7 PARTIAL / 2 FAIL / 2 BLOCKED #9334 has bug + regression, QA run · records-forms (FULL area) · e4e5c6e3 · 2026-08-17 · 2 PASS / 10 PARTIAL / 1 FAIL #9351 has bug (its 2026-08-17 extraction comment says the label moved to 【缺陷】被 multiple:true lookup 指向的对象 REST DELETE 全 400——cascadeDeleteRelations 依赖探针用裸等值查 JSON 列 #9362, but the removal never happened), and ten of the fourteen carry tracking. I made no label writes — none was in this card's mandate, and all fourteen are now closed so no sweep can pick them up. Cheap to fix if you want the record clean.
2. The checklist-accuracy corrections are not merely parked — they have a card, docs(qa): land the #9296 wave's environment facts in RUNNER.md and the 34 checklist-accuracy corrections in areas/*.json #9386, whose PR docs(qa): land the five re-verified environment facts in RUNNER.md #9427 (RUNNER environment facts) is merged and whose PR docs(qa): land the #9296 wave's 32 checklist item corrections and the console-session-auth environment fact #9475 (the 32 item corrections, including the no-active-org-session-semantics blocked marker and the first-run-loop un-blocking) is still open. My close-out comments call this "the ledger-edit PR"; docs(qa): land the #9296 wave's environment facts in RUNNER.md and the 34 checklist-accuracy corrections in areas/*.json #9386 / docs(qa): land the #9296 wave's 32 checklist item corrections and the console-session-auth environment fact #9475 is its name.
3. QA run · tier2a:api-cli-build (23/23) · e4e5c6e3 · 2026-08-17 · 15 PASS / 6 PARTIAL / 1 FAIL #9417's four fixture/environment findings designated for the fixture card Land the showcase seed fixtures that unblock the checklist's stock-fixture gaps (loginable 2nd member, publicSharing object, writable summary field, readable:false FLS grant) #9308 (absent plugin-reports, three inert sharing rules, showcase_declarative_connector_ping on a non-3000 boot, the missing credentialed personas) were not posted to Land the showcase seed fixtures that unblock the checklist's stock-fixture gaps (loginable 2nd member, publicSharing object, writable summary field, readable:false FLS grant) #9308 — outside this card's scope. They stay readable on the records and parked on tracking: full platform-checklist regression against main @ e4e5c6e3 — five-round orchestration #9296.
4. QA wave R0: provision the shared QA tree and take the machine-lane baseline at e4e5c6e3 #9298 (R0) is still open and is not a qa-run record, so it was out of scope here.

Nothing was left unclassified: every FAIL in the four completed runs had sufficient evidence on its own record to classify.

---

Generated by Claude Code

[os-zhuang]

ACCEPT & close — reviewer of record: skills seat, session session_01Rn7aaamsR99FXRqLcpL99q.

Verified against GitHub: label:qa-run is:open returns 0 (was 14); spot-checked #9333's close-out comment — per-item supersede verification against the tier-run union, carried-forward verdicts stated explicitly (the P2-unpinned Tier-3 hole named rather than silently dropped), machine-greppable lines present. Nine FAIL clauses dispositioned, all class (a); the two access-control FAILs recorded existence-only per the disclosure ruling. Two new defect cards filed (#9487 UNAUTHENTICATED envelope missing code; #9488 read-door 200 on unknown meta type) — both bare, in triage. Dup rule honored (#9454/#9468/#9469/#9362/#9391 linked, not re-filed).

Leftovers noted from the report, dispositioned here: label drift on now-closed records — harmless (closed records are outside every sweep), not worth writes; the checklist-accuracy corrections have a live home (card #9386, open PR #9475) — the anchor's parked list should be read together with that PR; #9417's fixture findings stay parked on #9296; #9298 (R0) is not a qa-run record — out of scope, correctly untouched.

Close-out debt: 14 → 0. Going forward the close-out duty lands in the skill text via #9484 so the debt never re-accumulates.

---

Generated by Claude Code