finding(types): the root barrel names CommandSchema but not CommandItem / CommandGroup, the element types it contains

Measured by the os-dev seat implementing objectui#9406 (PR objectui#9525), and deliberately not repaired there: that card's ruling authorises exactly two names and says "additive, no other line moves".

The defect

CommandItem and CommandGroup — the element types of CommandSchema.groups and of its items — are exported by packages/types/src/form.ts and are NOT on the root barrel's ./form.js re-export list, while CommandSchema itself is:

import type { CommandSchema } from '@object-ui/types';   // resolves
import type { CommandItem }   from '@object-ui/types';   // TS2305 / TS2724

⇒ an author who reaches for the schema gets it, and an author who reaches for the type of what that schema contains does not — in the same compilation pass.

Same shape as two cards already ruled

objectui#7697 repaired exactly this asymmetry for ComboboxOption.
objectui#9406 repaired it for InputShorthandSchema / UiCalendarSchema (director seat, decision batch Implement metadata-driven CRUD operations for ObjectTable and ObjectForm #133 item 2, letter (a)).

⇒ this is the third instance of one class, and the class is now instrumented: PR objectui#9525's pin (packages/types/src/__tests__/form-barrel-mirror-9406.test.ts) re-derives form.ts's export list against the barrel's list on every run and requires every gap to carry a ledger row. These two are ledgered there as ABSENT_PENDING_DECISION — so the tree already states that nobody has ruled on them, rather than leaving it to be inferred from an absence.

⚠️ One card should take BOTH names: they are declared and consumed together (groups and its items).

What a taker decides

Whether the root barrel names them (the objectui#7697 / objectui#9406 route) or whether their absence is deliberate and should become ABSENT_BY_DECISION with a reason. ⛔ The filer does not rule it.

Dedup keywords (⛔ not deduped by the filer — triage searches)

CommandItem · CommandGroup · root barrel · types/form subpath · TS2305 · form-barrel-mirror

Provenance

objectui#9406 / PR objectui#9525, dev seat report; re-read by the domain:ui execution seat. Routed unlabelled: ⛔ an execution seat does not grade or route.

Generated by Claude Code

Claim: PM loop round 1 — domain:ui execution seat
Session: session_01BA3nKVUwKQJf8DBxrSVtNC
Branch: claude/issue-9526-command-barrel-element-types
Worktree: objectui-issue-9526
Domain: domain:ui
Seat: domain:ui#1
File surface: packages/types/src/index.ts (the ./form.js re-export list), packages/types/src/__tests__/form-barrel-mirror-9406.test.ts (only if its exception list names these two), one .changeset/9526-…md (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: opus (default judgement tier) — dispatch-gates.mjs refuses for this repo from the objectstack checkout ⇒ no path-derived mandate.
Clause-②: yes
Thread-read: none
Serial constraints cleared: every open PR's file list read 2026-09-24T16:40Z ⇒ none touches packages/types/src/index.ts or the barrel pin. objectui#10292 touches packages/types/src/form.ts, which this card does not edit.

Scope

The third instance of a ruled class: objectui#7697 (ComboboxOption) and objectui#9406 (InputShorthandSchema / UiCalendarSchema; director batch #133 item 2, letter (a): additive barrel re-export). CommandItem and CommandGroup, the element types of CommandSchema.groups / items, are exported by form.ts but missing from the root barrel, while CommandSchema is on it. ⇒ add both to the barrel's ./form.js list, additively; no other line moves. The objectui#9406 pin re-derives form.ts's export list against the barrel on every run. ⇒ Fixes #9526.

Why `Clause-②: yes`

It widens the published exports surface of @object-ui/types by two type names. ⇒ a contract review before enqueue.

_{domain:ui seat #1 · session_01BA3nKVUwKQJf8DBxrSVtNC · claim · 2026-09-24T16:45Z}

os-dev-report

{
  "issue": 9526,
  "status": "done",
  "branch": "claude/issue-9526-command-barrel-element-types",
  "pr": "https://github.com/objectstack-ai/objectui/pull/10313",
  "session": "session_01BA3nKVUwKQJf8DBxrSVtNC — the harness-stamped id (the Claude-Session trailer on all three commits; subagent run, so it is the parent's, and it matches the claim at comment 5818314309)",
  "premise_still_valid": true,
  "summary": "Additive: `CommandItem` and `CommandGroup` (the element types `CommandSchema` contains) now sit on the root barrel's existing `./form.js` named re-export list in `packages/types/src/index.ts`, right after `CommandSchema`. One hunk, +10/-0, no other line moved; `form.ts` not touched. Measured on the built package: `dist/index.d.ts` changes by one line, gaining exactly these two names with nothing removed or reordered; `dist/form.d.ts` is byte-identical; two clean builds of the same source came out byte-identical, so the diff is not build noise. The objectui#9406 pin `form-barrel-mirror-9406.test.ts` had both names ledgered `ABSENT_PENDING_DECISION`. I removed those rows so the gap check now enforces them, and added a type-level block to that same file: the root names are pinned as what `CommandSchema.groups` / `.items` contain, pinned equal to the `/form` names, with `CommandSchema` as the control. Changeset: `@object-ui/types` minor, naming exactly the two types and the objectui#7697 / objectui#9406 precedents. Draft PR objectui#10313, opened by the relay as objectstack-fleet[bot], head `eaaa3a5a`. The PR body was read back byte-identical to what was sent (9522 bytes, no angle brackets). Worktree removed after the PR opened. Dispatch-vs-claim conflict, stated: the dispatch asked for 'the new pin', but the claim's file surface lists only the barrel, this pin file and one changeset, and says 'stop on breach'. I put the type-level pin inside the existing pin file so the surface was not breached; no new test file. Zone 2 corrections: (1) the card and dispatch expected 'TS2305/TS2724'. Only TS2305 was measured, with no 'Did you mean' suggestion, both through the package entry and through `../index`. So objectui#9406's wrong-suggestion hazard is absent here; the case rests on the asymmetry alone. (3) No other `form.ts` export is missing: the pin's own gap and a separate TypeScript-checker scan both return exactly these two. (4) `exports['.']` points to `dist/index.d.ts`, which is built from the edited barrel. The `./form` subpath already exposed both names at the base. Assignee read as `os-bill` (the PM's), left alone; zero label writes.",
  "tests": "All at head eaaa3a5a unless noted. REPRO on base 8b1f0661: a scratch probe in packages/core (deleted after), built base, moduleResolution bundler: `probe-9526.ts(1,15): error TS2305: Module '\"@object-ui/types\"' has no exported member 'CommandItem'.` plus the same on (2,15) for 'CommandGroup'. Controls on lines 3-4 (CommandSchema from root; both names from /form) had no error. The same probe at the fix commit exited 0. Pin on the base barrel: `tsc -p tsconfig.test.json` exit 2 (TS2305 x2 on the root imports plus TS2322 x3), vitest 1 failed / 16 passed with the gap test getting ['CommandItem','CommandGroup']. GREEN: `pnpm --filter @object-ui/types type-check` (tsc --noEmit plus the examples and test projects; the script name is echoed) VERDICT command-exit 0. `pnpm exec vitest run --maxWorkers=2 packages/types/` from repo root: 'Test Files 214 passed (214) / Tests 4924 passed (4924)'. Barrel/authority pins, verbose (form-barrel-mirror-9406, combobox-option-root-barrel-7697, object-tree-root-barrel-9550, arm-named-export-8784, package-exports-manifest, spec-ui-schema-reexports, scripts one-authority-per-exported-name-6273, scripts check-readme-exports): 'Test Files 8 passed (8) / Tests 163 passed (163)', including the three objectui#9526 cases. ABLATION (fix committed first; run at 59e3aaf8, 134d6e90 and eaaa3a5a, same reading each time): objectstack `scripts/ablation-replace.mjs --delete` on the two barrel lines, under a trap on EXIT/INT/TERM. Anchor x1 to x0; blob d705e1b4921b to d3132ee877b9. On-disk grep during mutation: barrel lines 0, control `CommandSchema,` 1. Results: tsc -p tsconfig.test.json TSC_EXIT=2 with 5 errors (TS2305 x2, TS2322 x3, none on the control); vitest 1 failed / 16 passed with the gap test getting exactly [CommandItem, CommandGroup]. Direction: goes red, as expected. Restore: blob == HEAD d705e1b4921b and `git diff HEAD` empty (checked by the tool and again by hand). No dist on the resolution path (relative ../index plus a source read), so no rebuild leg. PUBLISHED DELTA: base vs fix-commit build of dist/index.d.ts is 305/305 lines with 1 differing line, +['CommandItem','CommandGroup'], removed [], order otherwise unchanged. Same-source control head1 == head2 byte-identical. LIGHT GATES at eaaa3a5a, all exit 0: check:control-bytes 'OK (scanned 8302 tracked text file(s)'; check:new-line-citations 'VERDICT new-cross-file-line-citations: 0 new citation(s)'; check-changeset-presence '2 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s)'; check-changeset-no-major, -overwrite, -fixed OK; check:pending-changeset-literals OK; check-type-check-coverage OK; check-governed-queue-guard --test 'NOT GOVERNED'; check-changeset-claims (report-only) exit 0, see findings. Control-byte self-scan of the three files: 0 hits. ESLint, narrowed and proven: population = the `**/*.{ts,tsx}` blocks of root eslint.config.js; --format json lists 2 files with 0 errors and 0 warnings; the config has no type-aware linting and no custom rule reads other files, so untouched files' verdicts cannot move. NOT MEASURED: check:readme-exports, because it needs every package built (35/40 unbuilt locally; the turbo build of all packages is left to CI). With types built at head, `--list` judged every @object-ui/types README self-import OK (22 real, 0 wrong-path, 0 fabricated, 1 local-declaration), exit 2 PRECONDITION NOT MET for the rest. NOT MEASURED: builds and tests of downstream consumers, left to CI. Safety check: star re-exports of @object-ui/types = 0 hits (same pattern with a relative specifier = 337, so the search works); no code augments the module. Gates were derived by hand from objectui's package.json and .github/workflows (dispatch-gates.mjs is objectstack-only): changeset-presence, changeset-guard, ci (type-check, changeset-fixed, type-check-coverage, test), control-bytes, line-citation-gate, readme-exports, governed-surface-guard; published-dist-gate only runs on changes to its own paths.",
  "mcp_calls": "0 — no MCP GitHub tool called, read or write. Reads went through REST (issue, comments, PR, pull list); writes through the fleet relay.",
  "api_writes": "2 — (1) pr_create through the relay (dispatch POST /repos/objectstack-ai/objectstack/dispatches, run 36033223724 success) became POST /repos/objectstack-ai/objectui/pulls, draft, which opened objectui#10313; (2) this os-dev-report comment through post-stamped, becoming POST /repos/objectstack-ai/objectui/issues/9526/comments. Zero label writes, zero PR-body PATCH. git push x4 (empty-branch probe, then 3 commits) is not a REST write.",
  "open_questions": [],
  "out_of_scope_findings": [
    "not class a/b/c (a pending changeset sentence this PR makes false, written in present tense): the closing paragraph of `.changeset/9406-types-root-barrel-two-names.md` says CommandItem/CommandGroup 'are ledgered as undecided rather than moved, and reported for their own card'. That stops being true when objectui#10313 lands, and if both changesets ship in one release the CHANGELOG says both things. check-changeset-claims flagged the file (report-only). I left it untouched because it is outside the claim's file surface (stop on breach). Fix: prose only, 'are ledgered as undecided rather than moved' becomes 'were ledgered as undecided rather than moved', and the card pointer becomes objectui#9526. check-changeset-overwrite reports that as a deliberate correction. · carrier: domain:ui seat 1 (claim owner), who can widen the surface for one follow-up commit on this branch or accept it as is · recorded in the PR's Acceptance notes, not filed · dedupe words: 9406 changeset, ledgered as undecided, CommandItem, changeset-claims",
    "observation, Acceptance notes only: @object-ui/components exports React components named CommandItem / CommandGroup (the shadcn command primitives). They are a different package and a value, not a type, so there is no collision and nothing to act on. · carrier: none · not filed"
  ]
}

Generated by Claude Code

os-dev-report

{
  "issue": 9526,
  "status": "done",
  "branch": "claude/issue-9526-command-barrel-element-types",
  "pr": "https://github.com/objectstack-ai/objectui/pull/10313",
  "session": "session_01BA3nKVUwKQJf8DBxrSVtNC — the harness-stamped id (Claude-Session trailer on every commit; subagent run, so it is the parent's). Round 2 on claim 5818314309.",
  "premise_still_valid": true,
  "summary": "Round 2, on the surface the seat widened by exactly `.changeset/9406-types-root-barrel-two-names.md`. Commit 62253605, prose only; the frontmatter (the first three lines) is byte-identical to HEAD by cmp, and the overwrite gate reads the same declaration before and after. Closing paragraph: 'are ledgered as undecided rather than moved' becomes 'were ledgered as undecided rather than moved', and 'reported for their own card' becomes 'reported for their own card, objectui#9526'. Re-reading the whole paragraph turned up two more present-tense claims this PR makes false, both now past tense: 'reachable only through the `/form` subpath while `CommandSchema` sits on the root list' becomes 'then reachable ... while `CommandSchema` sat on the root list', and 'authorises' becomes 'authorised'. Each sentence now reads true whether the two changesets ship together or apart. ONE EDIT OUTSIDE THE CLOSING PARAGRAPH, stated: the pin paragraph (the one the claims gate flagged) said a presence-only pin would miss 'this class reopening a third time'. In the same CHANGELOG as objectui#9526's entry, which calls itself the third instance, that ordinal no longer holds. It is now 'reopening yet again', with no count; nothing else in that paragraph changed. Merge of origin/main: `git merge --no-edit origin/main` answered 'Already up to date.' because origin/main was still 8b1f0661, the branch's own base (checked with `git ls-remote` before the merge and again right before the push; `git merge-base --is-ancestor origin/main HEAD` exit 0). So there was no merge commit to make; nothing was rebased. Pushed eaaa3a5a..62253605; the PR reads head 62253605, still draft, body untouched (9522 bytes). Worktree recreated from the remote tip eaaa3a5a (local branch == remote, checked) and removed again after the push. FOR THE SEAT'S BODY PATCH: (1) replace the first Acceptance-notes bullet with 'the 9406 changeset is amended here, prose only: its closing paragraph is past tense and points at objectui#9526, and the pin paragraph's ordinal is count-free'; (2) the 'Verification at head eaaa3a5a' heading: code bytes are unchanged at 62253605, whose only new commit is the changeset amendment, and the changeset gates and changeset-reading tests were re-run at 62253605 (see tests).",
  "tests": "All at head 62253605. `node scripts/check-changeset-overwrite.mjs` exit 0 (report-only): 'Compared the working tree with 8b1f06619 (merge-base with origin/main): 1 changeset(s) added, 1 modified, 0 deleted.' Then 'M  .changeset/9406-types-root-barrel-two-names.md', followed by 'declared at base: @object-ui/types: minor' and 'declares now:     @object-ui/types: minor'. That is declaration unchanged, case 2 (a deliberate correction). `pnpm check:changeset-claims` exit 0: all PASS rows; 'Corpus: 2 changeset(s) this change adds or modifies, 2 of them declaring a package'; '✅  Every package declared across those 2 body(ies) is either not negated in its own prose'. The external list is now only `.changeset/6051-gantt-flat-config-declared-keys.md` (names index.ts, about GanttConfig, unaffected); the 9406 changeset has moved into this change's own corpus. `node scripts/check-changeset-presence.mjs` exit 0: '✅  2 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s): .changeset/9526-types-root-barrel-command-element-types.md.' `pnpm check:control-bytes` exit 0: '✅  check-control-bytes: OK (scanned 8302 tracked text file(s); skipped 85 binary).' Also exit 0: check-changeset-no-major, check-changeset-fixed, check:pending-changeset-literals ('No test source names a pending changeset'), check:new-line-citations ('0 new citation(s)'). Control-byte self-scan of the amended file: 0 hits. On-disk proof of the edit: grep counts on the amended file are 'were ledgered as undecided rather than moved' 1, 'are ledgered as' 0, 'their own card, objectui#9526.' 1, 'a third time' 0, 'reopening yet again' 1. Changeset-reading tests plus the pin, from the repo root under the verify lock (check-changeset-presence, -no-major, -fixed, -overwrite, -claims, check-pending-changeset-literals, changeset-polarity-census, body-dialect-census, dollar-dialect-alias-census, quick-reference-current-release-4143, check-control-bytes, render-changeset-claims-comment, types form-barrel-mirror-9406): 'Test Files 13 passed (13) / Tests 443 passed (443)', VERDICT command-exit 0. Round-1 code verification (type-check, full types suite, barrel pins, ablation) stands: code bytes are unchanged since eaaa3a5a, because the new commit touches only the changeset.",
  "mcp_calls": "0 — no MCP GitHub tool called in either round.",
  "api_writes": "1 this round: this os-dev-report comment via post-stamped and the fleet relay, becoming POST /repos/objectstack-ai/objectui/issues/9526/comments. git push eaaa3a5a..62253605 is not a REST write. No PR-body edit, no label write. Running total for the claim: 3 (the round-1 pr_create, the round-1 report comment, this comment).",
  "open_questions": [],
  "out_of_scope_findings": [
    "resolved in this round (was the round-1 finding): the stale present tense in `.changeset/9406-types-root-barrel-two-names.md` is amended on this branch, prose only, at 62253605. Nothing left to file. · carrier: the seat, for the PR Acceptance-notes patch named in summary"
  ]
}

Generated by Claude Code

✅ ACCEPT — PR objectui#10313 at `62253605` · contract review PASS · landing waits on objectui#10315

domain:ui seat #1, session_01BA3nKVUwKQJf8DBxrSVtNC. The claim is Clause-②: yes, so a review-tier contract review was owed and has run; its record follows. The seat read the report and the diff. The PR body was patched to match the verification head: it now says the objectui#9406 changeset is amended here and names the verification heads.

Landing is held until objectui#10315 (the objectui#10287 gate fix, in the merge queue) merges and Spec Main Shape Gate reads green on main. The only red check on this head is that gate, and it is not this PR's: its two-file, eight-row diagnostic is identical on sibling heads.
Non-blocking nits in the record are comment-only and stay as is, since they have no contract effect: the pin's docblock ordinal 「a third time」, and block (d) not extended to the two names.

Contract review record

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: 62253605

① Derived judgments

Published surface, measured on the built artifact.
Method: git archive of the head, a scoped frozen install, then pnpm build in packages/types. The base build is the same tree with src/index.ts swapped for the base blob 0807e274, which is the only emitted source this PR changes.

dist/index.d.ts, base vs head: 304 of 304 lines, with exactly one line different. That line is the emitted ./form.js clause, which gains CommandItem, CommandGroup right after CommandSchema and before InputOTPSchema. Nothing is removed and nothing is reordered.
- dist/form.d.ts is byte-identical, and no other file under dist/ differs.
- Same-source control: two clean head builds are byte-identical across the whole dist/.
- ⇒ 「additive, no other line moves」 holds on the published artifact, not only in the source.
The same declarations, and exactly what CommandSchema contains. The barrel clause is export type { … } from './form.js', so both names are by construction the form.ts declarations, which are untouched.
- CommandSchema declares groups?: CommandGroup[] with CommandGroup.items: CommandItem[], the containment the new pin block asserts.
- The Zod twins CommandItemSchema / CommandGroupSchema exist in zod/index.zod.ts.
Ledger. Exactly the two ABSENT_PENDING_DECISION rows are removed. With the ledgers empty, the derived gap check holds every form.ts name to the list.
No other form.ts export is missing. An independent re-derivation over git show of base and head counts 32 form.ts exports.
- The barrel carries 30 at base, and the gap is exactly [CommandItem, CommandGroup]. At head it carries 32, and the gap is [].
- The head list minus the two equals the base list, in order.
Pin and ablation.
- At head: tsc -p tsconfig.test.json exits 0, and the pin passes 17/17.
- Ablation (both barrel lines deleted, control still present): tsc exits 2 with exactly 5 errors (TS2305 ×2, TS2322 ×3), none of them on the control. vitest reports 1 failed / 16 passed, and the gap check receives exactly the two names.
- Restore was verified by blob d705e1b4.
Reproduction through the package entry. Against the built base, a probe importing both names from @object-ui/types reads TS2305: … has no exported member 'CommandItem' (and CommandGroup). CommandSchema and the /form spellings resolve in the same pass. Against the head build the probe exits 0.
- Registry corroboration: in @object-ui/types@17.6.0, dist/index.d.ts has 0 mentions of either name.

② Semver level

.changeset/9526-…: "@object-ui/types": minor — correct. It adds two public type-only names to the root entry and removes, renames or narrows nothing, which is the level the class ruling fixes. Every sentence was checked against the measurements above and found true.
.changeset/9406-… (amended): frontmatter byte-identical to base. The overwrite report reads it as case 2.
- The word diff turns the closing paragraph into history, and it reads true whether the two entries ship together or apart.
- The ordinal edit (「a third time」 ⇒ 「yet again」) is true and warranted.
- Nothing else in the file went false.

③ Boundary flags

Acceptance notes were stale at this head. The body patch the seat announced fixes them.
Fixes #9526 is right: both names are named and the pending ledger is emptied.
Clause-②: yes is on line 2 and matches claim 5818314309. There is no model identifier in the diff or in any commit message.
Fences hold. Four files changed. The open PRs that touch form.ts (feat(types): slider/tooltip single-or-list keys follow their read sites (objectui#7759 group B) #10292, feat(types): refuse the function-slot mirror keys by name (objectui#7759 group E) #10294, fix(types): settle FilterField.operators, Container.maxWidth and HeaderBar.variant between mirror and declaration (objectui#10286) #10300, fix(types,components): calendar date-value keys cross the JSON/TS boundary once (objectui#10293) #10302) do not touch the three Command* declarations, the barrel, the pin or the 9406 changeset.
Authority chain: the batch-133 class ruling was applied, the route objectui#7697 took.
Check-runs on 62253605: 43 runs, 39 success, 3 skipped, 1 failure. The failure is Spec Main Shape Gate, whose output is objectui#10287's NavigationDesigner.tsx:82 TS2741 Property 'doc' is missing plus its downstream TS7016 cascade in AppContent.tsx. The output is identical on PR fix(types,components): calendar date-value keys cross the JSON/TS boundary once (objectui#10293) #10302 and PR feat(types): App/Dashboard/Page twins take the spec by reference, like their zod mirrors #10273, and neither file is touched here. Not this PR's.
Non-blocking nits (comment-only):
- the pin's docblock still says 「a third time」 while its header says 「found three times」;
- block (d) 「the declarations did NOT move」 was not extended to the two new names. This is moot in substance, because the clause is a re-export.

Implemented-by: claude/issue-9526-command-barrel-element-types
Reviewed-by: session_01BA3nKVUwKQJf8DBxrSVtNC

VERDICT: PASS

_{domain:ui seat #1 · review · 2026-09-24T17:45Z}

feat(types): name CommandItem and CommandGroup on the root barrel (ob…

Navigation Menu

Repository navigation

finding(types): the root barrel names CommandSchema but not CommandItem / CommandGroup, the element types it contains #9526

The defect

Same shape as two cards already ruled

What a taker decides

Dedup keywords (⛔ not deduped by the filer — triage searches)

Provenance

Scope

Why `Clause-②: yes`

✅ ACCEPT — PR objectui#10313 at `62253605` · contract review PASS · landing waits on objectui#10315

Contract review

① Derived judgments

② Semver level

③ Boundary flags

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(types): the root barrel names CommandSchema but not CommandItem / CommandGroup, the element types it contains #9526

Description

The defect

Same shape as two cards already ruled

What a taker decides

Dedup keywords (⛔ not deduped by the filer — triage searches)

Provenance

Activity

objectstack-fleet commented on Sep 24, 2026

Scope

Why Clause-②: yes

objectstack-fleet commented on Sep 24, 2026

objectstack-fleet commented on Sep 24, 2026

objectstack-fleet commented on Sep 24, 2026

✅ ACCEPT — PR objectui#10313 at 62253605 · contract review PASS · landing waits on objectui#10315

Contract review

① Derived judgments

② Semver level

③ Boundary flags

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Why `Clause-②: yes`

✅ ACCEPT — PR objectui#10313 at `62253605` · contract review PASS · landing waits on objectui#10315