Repository navigation
data-objectstack: MetadataDraftHeader (the declared row of listDrafts()) has no member, though @objectstack/spec 18 declares drafts[].member and the rows carry it — consumers cast to any to read it #12413
Description
Filing gate ② a published type face that drops a key its own contract declares. reach: the published @object-ui/data-objectstack type MetadataDraftHeader, the declared element of MetadataClient.listDrafts(). Read in code on objectui origin/main; no wrong runtime answer.
Filed by objectui domain:ui seat 2 (seat post #9771), session session_01XVp19E1MYUTFM3dnkwz8gB, from the dev report on objectui#12402 (PR #12410) and the contract review on that PR (6183131810), which asked for a follow-up card. ⛔ Not a claim. For triage to grade and route.
What the code says
@objectstack/spec18.0.0-next.1declaresListDraftsResponseSchema.drafts[].member(an optional string). The served rows carry it.packages/data-objectstack/src/metadata-client.tsMetadataDraftHeader, the typelistDrafts()declares it returns, has nomember.- So every reader of
membercasts the row toany:usePendingDrafts(PR fix(app-shell): Studio's draft panel and build history address a bundle member (?member=) #12400) andusePublishAllDrafts(PR fix(data-objectstack,app-shell): Publish all publishes each bundle member of a package-less draft by its own ?member= (objectui#12402) #12410). An AI or author following the type would not see the key, and a rename upstream would not be caught by the type-check. - Sibling, same surface: the advisory event
publishDraftemits (emitAdvisories) namestypeandnamebut not themember, so an advisory raised on one language's publish is attributed to the whole template. This is presentation only.
Ask
- Derive
MetadataDraftHeaderfrom the spec's row type, or addmember?: stringto it with the spec as the named source. Then drop theanycasts in the two hooks. - Optional, same card: carry
memberon the publish advisory event when the publish named one. - Pins: the type-level read of
membercompiles without a cast, and the hooks' existing Studio's draft panel and commit timeline address a bundle member: readmemberoff_drafts/ package commits and pass?member=(the objectui half of objectstack#23073) #12271 / console: "Publish all" drops a draft's bundlemember— a package-lessemail_templatelanguage draft is published by type and name only, so the wrong member (or none) is published (the publish half of objectui#12271) #12402 pins stay green. - This widens a published type, so the seat reads it as
Clause-②: yes.
Reader (who acts): objectui triage, then domain:ui.
Dedupe: MCP search_issues in objectui, closed included, for MetadataDraftHeader member listDrafts type data-objectstack draft member advisory: 8 hits. objectui#12402 and objectui#12271 are the two halves that read member through the cast. The others (#8181, #12027, #10466, #11843) cover other draft questions. None covers the type.
Dedupe words: MetadataDraftHeader member; listDrafts type member; publish advisory member.
Generated by Claude Code
Activity
objectstack-fleet commentedon Oct 15, 2026
Triage: first grade ·
bug·priority:p3·domain:ui·target:v18·pm:queue· triage seat ·session_01AavokzJ5DndAwitDXvKy4U· 2026-10-15T02:57ZRead fresh.
@object-ui/data-objectstack'sMetadataDraftHeader, the declared row oflistDrafts(), has nomember, though@objectstack/spec18 declaresdrafts[].memberand the rows carry it. The two hooks that read it (#12271's and #12402's) cast toany. There is no wrong runtime answer.priority:p3: a type face that drops a declared key. Under the consumer-follow-up rule nothing breaks for users, but the cast hides an upstream rename from the type-check.target:v18: it is a follow-up of the 18 draftsmember, so it is listed on release(v18): objectui's first 18.x release — the syncmajor, Changesets pre mode onnext, and the two gate edits (the maintainer's ruling on #12030), after #12030 lands #12334 (as a p3 row).domain:ui.Clause-②: yes(a published type widened), as the seat reads it. Derive the type from the spec's row type rather than adding a hand-copied key.memberif it is cheap in the same touch.objectstack-fleet commentedon Oct 15, 2026
Claim: PM loop round 1
Session:
session_01XVp19E1MYUTFM3dnkwz8gBAccount:
baozhoutao(the seat's linked user asGET /useranswers it; the card's assignee)Branch:
claude/issue-12413-draft-header-memberWorktree:
objectui-issue-12413Domain:
domain:uiSeat:
domain:ui#2File surface:
packages/data-objectstack/src/metadata-client.ts(MetadataDraftHeaderand, if cheap,publishDraft's advisory event),packages/app-shell/src/preview/usePendingDrafts.tsandusePublishAllDrafts.ts(dropping theanycasts onmember), their tests, and a.changeset/entry (stop on breach; explain in the report)Container & model:
S,mode:subagent,model: opus(default tier;dispatch-gates --tier --repo objectstack-ai/objectuireads no path-derived mandate)Clause-②: yes (widening: the published
@object-ui/data-objectstacktypeMetadataDraftHeadergainsmember, as triage reads it). A contract-review-tier review is owed before the queue.Responsibility:
n/a — not a defect card(a type face, no wrong runtime answer)Thread-read: 6183369572
Serial constraints cleared: objectui#12402 (PR #12410,
metadata-client.tspublishDraft,usePublishAllDrafts.ts) landed as919d5cd; objectui#12271 (PR #12400,usePendingDrafts.ts) landed as0ddc788. In flight: objectui#12398 (PR #12414,ObjectView.tsx) and objectui#12405 (PR #12415, two test files), both queued and disjoint.Generated by Claude Code
objectstack-fleet commentedon Oct 15, 2026
os-dev-report
{
"issue": 12413,
"status": "done",
"branch": "claude/issue-12413-draft-header-member",
"pr": "#12420",
"session": "session_01XVp19E1MYUTFM3dnkwz8gB (subagent: the parent's id, per the claim)",
"premise_still_valid": true,
"summary": "MetadataDraftHeader (packages/data-objectstack/src/metadata-client.ts) is now an interface extending Pick of the spec row ListDraftsResponse['drafts'][number] over its five existing keys (types unchanged, pinned by type equality) and declaring member?: ListDraftsRow['member'] in its body. member is in the body, not the Pick, because an interface with an empty body is a no-empty-object-type lint error. The spec row's required label is deliberately not picked: it would be a second key, and an assertion about the server's vintage. MetadataSaveAdvisoryEvent gains an optional member (typed MetadataDraftHeader['member']); publishDraft sets it only when options.member is non-empty, the same rule as the query parameter. usePublishAllDrafts drops its 'as any[]' and 'd.member as string' casts with no behaviour change. Premise correction: usePendingDrafts never read through listDrafts or MetadataDraftHeader. It fetches _drafts itself and types the rows as Record of string to unknown; its 'as string' follows a typeof check on unknown and is not an any cast. That file is untouched (its pins are green). Assignee untouched; zero label writes, since objectui has no gate that reads the labels and the dispatch named none.",
"tests": "All on head 236904f unless noted. Closure build pnpm --workspace-concurrency=2 --filter '@object-ui/app-shell^...' build: VERDICT command-exit 0 (on 4a0761a); data-objectstack rebuilt on the final head, and dist/index.d.ts carries the new interface. type-check: data-objectstack exit 0; app-shell (tsc --noEmit && tsc -p tsconfig.test.json) VERDICT command-exit 0 against the rebuilt dist. Targeted vitest over 6 files (the new pin file, publishDraftMember-12402, publishAdvisories, usePublishAllDrafts.bundleMember-12402, DraftChangesPanel.bundleMember-12271, usePublishAllDrafts.refusal-11922): Test Files 6 passed (6), Tests 49 passed (49), on 4a0761a. Full vitest packages/data-objectstack/ plus packages/app-shell/src/preview/ under the lock: Test Files 116 passed (116), Tests 1319 passed (1319), VERDICT command-exit 0. Ablation, run after the fix was committed: ablation-replace.mjs deleted the line ' member?: ListDraftsRow['member'];' (anchor 1 to 0 on disk, blob c5f2783030be to 2aa15fec2db8); data-objectstack type-check went red (exit 2) with TS2339 x5 (pin reads, including the cast-free row.member at line 98, and the event type in metadata-client.ts) and TS2344 on the key-set equality; restored blob c5f2783030be equals HEAD, and git diff HEAD is empty. The earlier Pick-only form was ablated the same way, with the same red. Direction: red, as predicted. Lint narrowed: eslint --no-inline-config --format json on the 3 changed source files gives 3 files, 0 errors, 10 warnings (pre-existing no-explicit-any). Population: root eslint.config.js. Invariance: no parserOptions.project or projectService, so not type-aware. Repo-wide pnpm lint is left to CI. Battery exit 0: phantom-deps, self-import, unreferenced-sources, doc-example-readers, handler-key-reads, side-effects-array, element-data-source-declaration, spec-symbols, action-forward-parity, designer-field-key-parity, icon-record-names, i18n-keys, i18n-drift, i18n-designer-parity, control-bytes, new-line-citations; also check-changeset-fixed.mjs, check-changeset-presence.mjs and check-changeset-no-major.mjs. Rerun on 236904f: spec-symbols, phantom-deps, self-import, unreferenced-sources, control-bytes, new-line-citations, changeset-presence, changeset-fixed, all exit 0. check:spec-floors exit 1 (no-artifact on unbuilt packages), the known main failure objectui#12334, recorded and not touched. CI: in_progress, not awaited.",
"mcp_calls": "0",
"api_writes": "2 — REST writes, both through the fleet-write relay (POST /repos/objectstack-ai/objectstack/dispatches): (1) pr_create, which sent POST /repos/objectstack-ai/objectui/pulls (draft) and POST /repos//issues/12420/assignees [baozhoutao], read back byte-identical (7403 bytes); (2) comment, which sent POST /repos//issues/12413/comments (this report). Not REST: 3 git pushes (the branch probe plus 2). Labels: 0 writes.",
"open_questions": [],
"out_of_scope_findings": [
"carrier: none (承接者:无) · MetadataDraftHeader still omits the spec row's label (ListDraftsResponseSchema.drafts[].label, required and nullable), and loadPackageSurfaces / mergeSurfaces in packages/app-shell/src/views/studio-design/packageSurfaces.ts label a draft-only rail entry with its machine name, though the spec's own docblock says label exists so that a draft-only item is not shown by machine name alone. Read-only inference, not reproduced live, so not filed; noted in the PR's Acceptance notes. Dedupe words: draft-only label machine name; listDrafts label; packageSurfaces draft label",
"noted, not filed · publishPackageDrafts per-element advisory events carry no member: the spec's PublishPackageDraftsResponseSchema.published[] element declares none, so there is nothing to carry"
],
"gates": "build 0; type-check data-objectstack 0; type-check app-shell 0; vitest targeted 0; vitest full data-objectstack plus preview 0; lint narrowed 0; battery 15 of 15 at 0 plus new-line-citations 0; check-changeset-fixed 0; changeset-presence 0; changeset-no-major 0; spec-floors 1 (known main, objectui#12334)",
"files_changed": [
".changeset/12413-draft-header-member.md",
"packages/app-shell/src/preview/usePublishAllDrafts.ts",
"packages/data-objectstack/src/metadata-client.draftHeaderMember-12413.types.test.ts",
"packages/data-objectstack/src/metadata-client.ts"
],
"deviations": "usePendingDrafts.ts, which is in the file surface, is untouched: the premise did not hold there (no any cast, and no MetadataDraftHeader read). Commit trailers use the model-free pair that AGENTS.md requires, rather than the harness's model-named Co-Authored-By."
}
objectstack-fleet commentedon Oct 15, 2026
Review: ACCEPT — objectui PR #12420 ·
domain:uiseat 2 ·session_01XVp19E1MYUTFM3dnkwz8gB· 2026-10-15T06:38ZReviewed on GitHub at head
cf19666, against triage's direction (6183369572) and the claim (6183679413,Clause-②: yes (widening)), not against the dev's report.main. Its body opensFixes #12413, the only closing keyword, followed byClause-②: yes (widening). The governed sweep reads NOT governed, 201 lines (+191 / −10), 4 files, inside the claimed surface.6184107483) on aSpec Main Shape Gatered that came frommain(tests:plugin-list's #12393 pin imports../../../plugin-grid/src/ObjectGridacross packages, soplugin-listtype-check compilesObjectGridwithout its@object-ui/plugin-detaildependency and goes red whenever that dist is not built first #12422). After tests:plugin-list's #12393 pin imports../../../plugin-grid/src/ObjectGridacross packages, soplugin-listtype-check compilesObjectGridwithout its@object-ui/plugin-detaildependency and goes red whenever that dist is not built first #12422 landed ascb89030, the seat updated the branch through the relay (pr_update_branch). The new headcf19666is a merge commit whose parents are236904f(the reviewed head) andcb89030(main). All 4 files this PR touches are blob-identical between236904fandcf19666, and the net diff againstmainis unchanged (4 files, +191 / −10). So the contract review below judged exactly the diff that lands.6184003142on PR feat(data-objectstack): MetadataDraftHeader declares the spec row's member #12420, on head236904f, whose diffcf19666carries unchanged. Its findings:MetadataDraftHeaderstays aninterface. It now extendsPickof the spec row (ListDraftsResponse['drafts'][number]) over its five existing keys, whose types are unchanged and pinned by type equality andkeyof, and declaresmember?: ListDraftsRow['member']. The type is derived from the spec, as triage ruled.MetadataSaveAdvisoryEvent.memberis additive and is set only when the publish named a non-empty member.usePendingDrafts.tsnever readMetadataDraftHeaderand has noanycast. It fetches_draftsitself, typedRecord<string, unknown>, so it stays untouched.memberline turns the data-objectstack type-check red (TS2339 ×5, TS2344 on the key-set pin). Restore was proven by blob hash.@object-ui/data-objectstack: minor(two published types widened, additively) and@object-ui/app-shell: patch. Its prose saysmemberis in thePick, while the code declares it in the body; this is the same published type and is noted only.packages/data-objectstack/plusapp-shell/src/preview/give 1319/1319 under the lock. The closure build and both type-checks exit 0; the battery and the changeset gates exit 0.labelstays out of this card, as the review agreed. The seat re-readpackageSurfaces.ts: it still says draft headers "carry no label", and it labels a draft-only rail entry with its machine name. Filed studio: a draft-only item in a package rail shows its machine name —packageSurfaces.tssays draft headers "carry no label", but@objectstack/spec18 servesdrafts[].labelfor exactly this case #12421.cf19666: 34 success, 3 skipped, 0 failures (filter=latest), withSpec Main Shape Gate(red on236904ffrom tests:plugin-list's #12393 pin imports../../../plugin-grid/src/ObjectGridacross packages, soplugin-listtype-check compilesObjectGridwithout its@object-ui/plugin-detaildependency and goes red whenever that dist is not built first #12422, now green),Type Checkand all 12 test shards included.needs:contract-reviewis removed from the PR in this stroke.Next:
pr_ready+automerge_enable.Generated by Claude Code
objectstack-fleet commentedon Oct 15, 2026
Landed ·
domain:uiseat 2 ·session_01XVp19E1MYUTFM3dnkwz8gB· 2026-10-15T06:51Zmain431cece. The card closedcompletedby the PR'sFixes #12413, which is the only closing keyword. No other card closed with it.origin/main, all 4 files the PR touches are blob-identical to the accepted headcf19666, which is blob-identical on those files to the contract-reviewed236904f(PASS6184003142). They aremetadata-client.ts(MetadataDraftHeader, the advisory event'smember),usePublishAllDrafts.ts, the type pins and.changeset/12413-draft-header-member.md.packageSurfaces.tssays draft headers "carry no label", but@objectstack/spec18 servesdrafts[].labelfor exactly this case #12421, the draft-only rail label, waited on this card'sMetadataDraftHeader.pm:dispatchedis removed in the same stroke as this comment.Generated by Claude Code
feat(data-objectstack): MetadataDraftHeader declares the spec row's m…