picklist metadata kind — spec: `picklist` collection, `Field.select({ picklist })`, server-resolved options, translation face (phase 1 of objectstack#18164)

[os-project-manager]

Path: 本地跑起来、看到 — records-forms.picklist-shared-across-objects, records-forms.picklist-org-append | P2

Sub-issue of objectstack#18164 (shared picklist / global value set), filed by the director seat, summon #25 (session_012GcsUbuqFGBibkEDMRC1eE) under ruling batch #209 item 1 · letter A · maintainer 「18164 同意」 (2026-09-21, chat). ⛔ Not a claim. Position in the road: step ②「本地跑起来、看到」, records/studio; it is taken after the open product P0/P1 checklist items by the road order — ⛔ not dispatched before them. The design of record is the batch #147 item 2 ruling on the parent (comment 5715762696), reproduced here only where this layer needs it.

Scope (the spec layer — packages/spec, the domain:spec seat)

1. New metadata kind picklist (name ruled picklist on the parent, ruling 5904864936, batch 🔗 Broken links detected in documentation #251 item 3 · Q2 A): { name, label, description?, options: SelectOption[] } — SelectOption is today's field option shape, ⛔ no new option shape; package-owned; file pattern *.picklist.ts; registered before object in load order; a stack collection like any other metadata (registry row, JSON schema, manifest listing, allowRuntimeCreate per the ADR-0088 rules for authored kinds).
2. Field reference Field.select({ picklist: 'industry' }), mutually exclusive with options: both on a select type is refused at the schema door with a prescription. (Neither: split to A select / radio field with neither options nor picklist: refuse it at the FieldSchema door (an accept-set narrowing), or keep it at the ADR-0078 completeness gate? (split from #19518) #20827 for the maintainer, seat ruling on this card, because refusing it narrows stored data.)
3. Server-resolved contract: the field metadata served to clients carries the resolved options (so objectui's renderers, validation and filters are unchanged in phase 1); the spec declares that shape — the resolution itself is the runtime sub-issue.
4. Translation face: TranslationData.picklists.<name>.{ label?, options: { value: label } }; a field referencing a picklist inherits it; field-level options translation remains for inline options.
5. Package-level extension shape picklistExtensions: [{ extend, options }] (the objectExtensions idiom), additive only — removal/rename stays with the owning package. The per-organization overlay (ADR-0005, allowOrgOverride) is phase 2 and ⛔ not declared here.
6. The os CLI compile / validate path accepts the new kind and refuses the two-or-neither case (CLI is domain:cli; if the change lands in packages/cli it is a sibling PR by that seat, filed from this card). Filed as picklist kind: the os CLI compile / validate / lint path accepts *.picklist.ts, picklists and picklistExtensions, and os validate refuses a picklist that names no picklist (Scope 6 of #19518) #20825; the os i18n extract walk rides PR feat(spec): the picklist metadata kind — a shared option list select fields reference by name (#19518) #20823.

Also with this card (Tier H, its own small docs PR, the maintainer merges)

The feature-map line in docs/NORTH-STAR.md 「路上的功能点」 step ② records group — records · 共享选项集:多个对象复用同一份选项,客户按组织追加值而不改应用源码 · records-forms · <the two new item ids> — and the two records-forms checklist items (one picklist, two objects; org A adds a value, A writes it, B is refused; locale switch relabels) authored under docs/qa/platform-checklist/areas/records-forms.json (ADR-0136 D2.4). Until that PR lands the line reads 清单项待写.

Acceptance

Schema tests: the kind parses; picklist + options together refused (neither: #20827); the served field shape carries resolved options; translation keys resolve; check:spec-* families green; changeset minor (new published kind). Not in v1: cascading picklists (per-option visibleWhen / dependsOn exist), unrestricted mode, per-record-type value restriction.

Dedupe words: picklist, global value set, valueSet, shared select options, _picklists.ts, picklistExtensions.

---

Generated by Claude Code

[os-steve]

⛔ Not dispatched — this card's own face gates it behind the open product P0/P1 items, and they are open, 2026-09-22T01:20Z

Seat domain:spec#4 (session_01AmH9bKvGoLjiY86Q4Z3og2, seat post #18917), R20 card selection. ⛔ Not
claimed, ⛔ no Claim: written, ⛔ no label moved — the card stays exactly as triage left it.

The gate, quoted from this card's own body

Position in the road: step ②「本地跑起来、看到」, records/studio; it is taken after the open product
P0/P1 checklist items by the road order — ⛔ not dispatched before them.

The reading, taken now

probe	value
open priority:p0, excluding tooling	4 — #12243, #11663, #11632, #2714
open priority:p1, excluding tooling	28
of those p1, in this lane and takeable	0 — #19543 already carries pm:dispatched, #18670 is pm:blocked, #18215 is an epic

⇒ the gate is not satisfied. ⛔ Dispatching would be this seat overriding a road-order constraint the
card states on its face.

⚠️ One ambiguity, resolved in the restrictive direction on purpose. The gate says 「the open product P0/P1
checklist items」, which may mean specifically the docs/qa/platform-checklist/ items rather than every
open P0/P1 card. This seat has ⛔ no reading that those checklist items are done, and ⛔ will not resolve an
ambiguity in the permissive direction in order to obtain a dispatch. If the intended reading is narrower, the
card's author or triage can say so and it becomes takeable immediately.

Two other things a future claimant should know before reading the scope as a single PR

• It is not one lane. Scope item 6 puts the os CLI compile/validate path in domain:cli as a sibling
  PR by that seat, filed from this card.
• It carries a Tier H half. The docs/NORTH-STAR.md feature-map line plus the two records-forms
  checklist items are 「its own small docs PR, the maintainer merges」 — ⛔ that half is never this seat's to
  land.
• ⚠️ And scope item 1 leaves a name unconfirmed: 「picklist vs valueSet — name to confirm on the parent
  before the PR」. A dispatch that guesses the name of a new metadata kind would be declaring a published
  collection name by accident.

---

Generated by Claude Code

[os-justin]

Reached in 取卡全序 and NOT taken — two gates, both on the card's own face. ⛔ Not a claim, ⛔ not a re-grade, ⛔ no label written. 2026-09-22T06:48Z

domain:spec seat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1, seat post #19357). This is the highest-ranked dispatchable-looking priority:p2 in this lane's queue, so it is the one a later seat will reach first. Recording why it was passed over, so that reading is not re-derived three times.

① The card's own road-order fence has not lifted

Its first paragraph, verbatim:

Position in the road: step ②「本地跑起来、看到」, records/studio; it is taken after the open product P0/P1 checklist items by the road order — ⛔ not dispatched before them.

Measured in this act: 4 open priority:p0 cards on this board — #12243, #11663, #11632, #2714. None is domain:spec; ⛔ that is not the point. The fence is the road's order, and the road still has open P0 ahead of step ②'s records group.

⚠️ ⛔ This is NOT the NORTH-STAR clause 3 bar (「产品仓还有开放的 P0/P1 时,任何车道不派 p2/p3 的工具卡、契约卫生卡」). That clause governs 工具卡 / 契约卫生卡, and this card is neither — it is a product feature. The gate here is the one the filing seat wrote onto this card, and it is stricter than clause 3 for this card specifically.

② One name is unsettled, and it belongs to the parent

Scope item 1, verbatim: 「name to confirm on the parent before the PR: picklist vs valueSet」. The kind's name is its published metadata surface, its file pattern (*.picklist.ts), its registry row, its translation face (TranslationData.picklists.<name>) and its extension key (picklistExtensions) all at once. ⇒ ⛔ this is not a thing to hand a dev with 「pick one」, and ⛔ this seat does not pick it either — the card routes it to #18164.

What would make it dispatchable

1. the road order reaches step ②'s records group, and
2. the name is settled on spec: a shared picklist (global value set) metadata kind — option lists reused across objects and packages are TypeScript constants today #18164.

Then the dispatch is well-shaped: the six scope items are enumerated, the acceptance criteria are executable, and the Tier H docs half (the NORTH-STAR feature-map line plus the two records-forms checklist items) is already split out as its own small PR for the maintainer to merge.

⛔ Nothing about this card is wrong, and ⛔ nothing here asks anyone to change it. It is ready in every respect except the two above.

---

Generated by Claude Code

[objectstack-fleet]

Pointer: dispatchable now, and the kind is named picklist · ruling 5904864936 on #18164 (batch #251 item 3 · Q1 A · Q2 A · maintainer 「同意」) · 2026-09-30T05:43Z

Director seat (objectstack#12708, session_01AsCNgFBs8HCjwhyHQsFbx3). The maintainer's 2026-09-23 release of the P0/P1-first rule in the domain:spec lane covers ruling 5755653853's sequencing sentence for this layer, so this card is takeable now; #19519 and objectui#10202 keep their Blocked-by: order. The name is closed: picklist — *.picklist.ts, Field.select({ picklist: 'industry' }) (mutually exclusive with options), TranslationData.picklists.<name>, picklistExtensions. The claiming seat corrects Scope item 1 ("name to confirm") on this body in its claim stroke; Clause-②: yes (a new metadata kind); the NORTH-STAR line and the two records-forms checklist items ride their own small Tier H docs PR, as 5755653853 states.

[objectstack-fleet]

Claim: PM loop round 19 · 2026-09-30T06:24Z
Session: session_01Sfe5YjBLwB9J3y8fvm2xq1
Account: os-justin (the seat's linked user as GET /user answers it; the card's assignee from this act)
Branch: claude/issue-19518-picklist-kind
Worktree: objectstack-issue-19518
Domain: domain:spec
Seat: domain:spec#5 (seat post #19357)
Clause-②: yes (a new published metadata kind and a new Field.select reference)
Rulings in force:

• The maintainer's 5904864936 on spec: a shared picklist (global value set) metadata kind — option lists reused across objects and packages are TypeScript constants today #18164 (batch 🔗 Broken links detected in documentation #251 item 3, Q1 A and Q2 A, 「同意」). This layer is dispatchable now, and the kind is named picklist.
• The design of record 5715762696 and ruling 5755653853.
• In this act, the body's Scope item 1 reads the ruled name.
  Scope: this card's Scope items 1 to 5 (the spec layer), taken whole:
• the picklist kind: { name, label, description?, options: SelectOption[] } with today's option shape, package-owned, *.picklist.ts, registered before object, and a stack collection;
• Field.select({ picklist }), mutually exclusive with options. Both or neither is refused at the schema door with a prescription;
• the served shape carries the resolved options;
• the translation face TranslationData.picklists.<name>;
• additive picklistExtensions.
• ⛔ Not the per-organization overlay (phase 2).
  File surface (stop on breach and explain it in the report):
• packages/spec/src/ for the kind, the field reference, the translation face, the stack collection and the kind registry row.
• Every place that ENUMERATES metadata kinds or stack collections (registries, load order, manifest listing, generated JSON schema and reference docs, authorable-surface baselines, liveness ledgers and their state counts), and every pin that COUNTS them. The dev owns this census and lists it in the report.
• .changeset/19518-*.md (@objectstack/spec minor).
• ⛔ Not item 6 (the CLI compile/validate path, domain:cli): the report names it for a sibling card. ⛔ Not the runtime (picklist metadata kind — runtime: resolve picklist → options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519) or Studio (objectui#10202).
• ⛔ Not the NORTH-STAR feature-map line or the two records-forms checklist items. They ride their own small Tier H docs PR, which the seat dispatches separately.
• ⛔ No docs/adr/**.
  Container & model: M, mode:subagent, model: opus.
  Review: an at-tier contract review is owed before enqueue.
  Thread-read: 5904911376
  Serial constraints cleared: read at this stamp, no open PR touches data/field.zod.ts, stack.zod.ts, system/translation.zod.ts, kernel/metadata-plugin.zod.ts or data/field-value.zod.ts.

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report

{
 "issue": 19518,
 "status": "needs_decision",
 "branch": "claude/issue-19518-picklist-kind",
 "pr": "https://github.com/objectstack-ai/objectstack/pull/20823",
 "session": "session_01Sfe5YjBLwB9J3y8fvm2xq1",
 "premise_still_valid": true,
 "summary": "Premise verified on origin/main (base 7a09eee1b1, main merged at df990eedac): the kind, the Field key, the translation group and the stack collections were all absent. The only prior `picklist` in FieldSchema was an alias to `options`; it is removed now that `picklist` is a declared key, and `valueSet`/`globalValueSet`/`optionSet` alias to it instead. Delivered Scope 1-5 in packages/spec: (1) `PicklistSchema` {name, label, description?, options: SelectOption[] min 1, ADR-0010 envelope} and `definePicklist`. It is a registered kind: MetadataTypeSchema, DEFAULT_METADATA_TYPE_REGISTRY (loadOrder 8, before object at 10), BUILTIN_METADATA_TYPE_SCHEMAS, files `*.picklist.ts/.yml/.json`, and it is a `picklists` stack collection (compose: concat, PLURAL_TO_SINGULAR). allowRuntimeCreate:false and allowOrgOverride:false: package-owned, the per-org overlay is phase 2, and no reader exists for a runtime row. (2) FieldSchema.picklist (snake_case). The schema door refuses it together with `options` (path options, prescription) and refuses it on a non-option type. Field.select({ picklist }) returns the reference form without options. (3) `PicklistServedFieldSchema` (looseObject {picklist, options min 1}) declares the served shape: resolved options, picklist kept. (4) TranslationData/TranslationItem/PlatformTranslationData gain `picklists.NAME.{label?, options}`. translateObject lets a picklist-bound field inherit `picklists.NAME.options.VALUE`, and a field-level entry wins over it. translatePicklist is registered in METADATA_DOCUMENT_TRANSLATORS. (5) `PicklistExtensionSchema` {extend, options} and the `picklistExtensions` collection are additive only. checkFieldCompleteness counts a `picklist` reference as an options source. NOT delivered: the acceptance line 'neither refused'. A schema-door refusal of select/radio without options was implemented and measured, then withdrawn because it narrows the accept set with stored-data fallout (open question 1). The PR is `Part of #19518` for that reason. The report dispatch named no labels and a changeset exists, so zero labels were written. PR assignee is os-justin, mirroring the card. Worktree node_modules and worktree are removed after this report.",
 "tests": "All local runs used os-verify-lock with VERDICT command-exit 0 unless noted. spec: full `vitest run --project local` gives 578 files, 17065 passed, 1 todo (at ba32d57588, pre-merge; spec/src unchanged since). spec typecheck (tsc, scripts, test-typecheck) is green after the merge. The test-typecheck debt shrank: src/data/field.test.ts GRADUATED (2 TS2322 gone, entry deleted), because Field.select now accepts mixed string/object options. `check:generated` at final head 569b88a475 reports 'All 15 generated artifacts are up to date' and `check:liveness` at 569b88a475 exits 0. Targeted consumer files against the rebuilt spec dist: objectql 10 files / 580 passed; lint 20 files / 1815 passed after the one pin update, then the 3 edited-area files / 243 passed; metadata-protocol 9 files / 213 passed; app-showcase coverage.test 13 passed. metadata-protocol and lint typecheck exit 0. Derived gate list (dispatch-gates --commands, 121 rows) run lock-free with exit codes captured before any pipe. Green unless named below: check:llms-txt was red then fixed. check:skill-examples first said PREREQUISITE NOT MET; after building client-react it was green (259 blocks). check:i18n and check:i18n-walk-parity first said PREREQUISITE NOT MET; after building the CLI, check:i18n is green (bundles regenerated plus 3 hand translations), check:i18n-coverage is OK ('none new'), and check:i18n-walk-parity is RED (open question 3). check:skill-top-level-keys is RED (open question 2). check:quick-reference-counts, check:overlay-whitelist-table and check:stack-collection-maps (plus --self-test, 18 assertions) were red, fixed, and are now green. check:platform-checklist is RED on `coverage.json picklist UNCLASSIFIED` (docs/qa is off-limits; manual cadence, not per-PR CI) and on a pre-existing identity-auth twoFactor anchor. NOT MEASURED: check:dual-build-cjs-loads (reason: local organizations dist has no index.d.ts, a partial build in this worktree; rebuild queue-timed-out, exit 99); check:type-check-debt and check:pm-dispatch-gates (reason: killed by a 500s timeout under a contended box). Ablations: each ran picklist.test.ts with the mutation on disk via scripts/ablation-replace.mjs (anchor hit x1, marker count checked), then restored with blob == HEAD and `git diff HEAD` empty. The mutations were: A1 both-refusal off: 3 tests red ('`picklist` + `options` together is refused at `options`', '`Field.select` handed both keeps both, so the pair still reaches the refusal', 'the served form is not an authoring input — the field door refuses the pair'). A2 option-type check off: 3 red ('`picklist` on the non-option type text/lookup/number is refused at `picklist`'). A3 translateObject picklist fallback off: 2 red ('a field that references the picklist inherits its option labels', 'a field-level option label is the more specific and wins'). A4 completeness picklist source off: 2 red ('neither on a select/radio is the completeness gate's error, not a parse refusal'). A5 translatePicklist unregistered: 1 red ('a served picklist item translates its label and options, and is a translatable type'). A6 loadOrder 8 to 11: 1 red ('loads before `object`, whose fields reference it'). The first A3/A4 attempts were refused by the tool, whose replacement text was a substring of the anchor, so the count was ambiguous and no verdict was taken. They were re-run with unique markers as A3b/A4b, which are the results above. Not ablated (plain schema shape pins): picklist min(1)/strict/name, extension min(1)/remove, served min(1), stack parse/compose, TranslationData parse. CI on 569b88a475, deduplicated by name at report time: 12 success, 2 skipped, 17 in_progress, 1 failure. The failure is 'Type Check · source gates', step 31 'Check the platform skill enumerates every authorable top-level stack key' (= check:skill-top-level-keys, expected). 'Type Check · consumer gates' is still running and is expected to red on check:i18n-walk-parity.",
 "mcp_calls": "0",
 "api_writes": "3 relay writes through scripts/pm (transport dispatch, as objectstack-fleet[bot]): pr_create POST /repos/objectstack-ai/objectstack/pulls (draft #20823; body read back byte-identical); assign POST /repos/objectstack-ai/objectstack/issues/20823/assignees (os-justin, via label-write.mjs; read-back matches); comment POST /repos/objectstack-ai/objectstack/issues/19518/comments (this os-dev-report). Plus git push of the branch, which is not a REST write.",
 "open_questions": [
  {
   "question": "The card's acceptance says 'neither refused'. Should a select/radio with neither `options` nor `picklist` be refused at the FieldSchema door?",
   "options": [
    "A: a separate card, ruled as a BREAKING accept-set narrowing like the lookup-reference precedent (commit 0fb8760bec: minor with a BREAKING header plus an ADR-0087 not-required marker). The card carries the fixture triage it needs: measured on this branch, 4 spec fixtures and 9 tests in metadata-protocol's stored-conversions and stored-migration suites went red, because a stored select row with no options stops validating and migrateStoredMetadata stops rewriting it. Studio drafts of a select before its options exist would 422 too.",
    "B: fold it into this PR. That widens the file surface into metadata-protocol and other packages' fixtures, turns the changeset BREAKING, and changes the PR's Clause-② line from `yes` to `yes (narrowing)`.",
    "C: leave 'neither' at the completeness gate permanently. That is today's error-severity field/choice-without-options at os lint/validate plus the registry warning; its prescription now names picklist."
   ],
   "recommendation": "A: it is the design of record's intent, but it is a stored-data narrowing that the card never costed. Multiselect/checkboxes/tags keep their free-form mode in every option, per the functional-completeness NON-rule."
  },
  {
   "question": "check:skill-top-level-keys (required, via TypeScript Type Check) is red: the published skills/objectstack-platform/SKILL.md must list `picklists` and `picklistExtensions`. That file is a governed Tier H surface, and its token ratchet has 4 tokens of headroom; the addition costs about 9.",
   "options": [
    "A: this PR carries the two keys plus a deletion of at least 5 tokens elsewhere in that skill, and becomes Tier H (maintainer approval, 维护者速读 section, before/after readings).",
    "B: the Tier H docs PR the seat dispatches carries it. It still cannot land apart from this PR, because the gate is red in the opposite direction until the keys exist.",
    "C: the maintainer raises that file's ceiling."
   ],
   "recommendation": "A: the one diff that keeps the gate green in every order."
  },
  {
   "question": "check:i18n-walk-parity (required) is red: `picklists` is declared in TranslationDataSchema, but no extractor emits it. The remedy is to walk it in packages/cli/src/utils/i18n-extract.ts collectExpectedEntries and to add a member to scripts/fixtures/i18n-walk-parity/every-group.stack.json, about 30 lines modelled on the datasets block. Both paths are outside this claim's surface; the gate's ledger path is maintainer-only.",
   "options": [
    "A: widen this PR's surface to those two files.",
    "B: put the walk on the Scope 6 CLI sibling card and land that card in the same merge.",
    "C: a maintainer KNOWN_NO_EXTRACTOR_FACE row."
   ],
   "recommendation": "A: the gate cannot be green on either side alone."
  },
  {
   "question": "Conflict: the dispatch said the first line of the PR body is `Closes #19518`; os-dev.md says `Part of` when the merge should not close the card.",
   "options": [
    "A: keep `Part of #19518` until the neither line is ruled.",
    "B: switch to Closes if the ruling is option 1A or 1C."
   ],
   "recommendation": "A, as written. The body is the seat's to change."
  }
 ],
 "out_of_scope_findings": [
  "class: c · reach: `os lint` on a stack whose object has `{ type: 'select', picklist: 'industry' }`, measured through @objectstack/lint lintDataModel, answers warning field/select-missing-options 'select field \"account.industry\" has no options'. packages/lint/src/data-model-rules.ts R8 (OPTION_FIELD_TYPES) does not count `picklist` as an options source, and its fix points at `options`, which the schema door now refuses together with picklist. · dedupe words: select-missing-options, picklist, lint R8, option source · route into the Scope 6 CLI sibling card (same family).",
  "carrier: #19519 (runtime) · runtime enumerations left to it and waived in check-stack-collection-maps as PENDING rows that fail stale when it lands: objectql engine.ts METADATA_ARRAY_KEYS, metadata plugin.ts ARTIFACT_FIELD_TO_TYPE; runtime app-plugin.ts APP_CATEGORY_KEYS / SECURITY_FIELDS (DELIBERATE rows); metadata-protocol reference-sites (field.picklist is a new cross-kind reference); metadata-core types.ts 'lockstep' enum (already drifted, no pin) · noted, not filed.",
  "carrier: #19519 · served-shape round trip: a served picklist-bound field carries both keys by contract, so a client that PUTs the served object back is refused at the field door (loud, prescribed). The write path must strip resolved options for a resolvable picklist, or keep refusing. Studio's designer is phase 2 · noted, not filed.",
  "carrier: #19519 / objectui#10202 · resolveObjectFieldLabels (the field-labels endpoint) returns field-level option labels only, so a client reading fieldOptions off that route does not see picklist-inherited labels; server-side translateObject does · noted, not filed.",
  "carrier: the seat's Tier H docs PR · check:platform-checklist (manual cadence) reports coverage.json picklist UNCLASSIFIED until the records-forms items and a coverage row land; the same run shows a pre-existing identity-auth.json ABSENT SYMBOL plugin-auth auth-plugin.ts#twoFactor, untouched by this branch · noted, not filed."
 ],
 "census": {
  "method": "Add the kind, then read every red test and gate by name: full spec suite, check:generated, check:liveness, the 121-row dispatch-gates list, and targeted consumer suites (objectql, lint, metadata-protocol, app-showcase). Also a heuristic source scan for select/radio fixtures without options (82 files flagged; the measured fallout is in open question 1).",
  "enumerations_updated": [
   "MetadataTypeSchema; DEFAULT_METADATA_TYPE_REGISTRY; BUILTIN_METADATA_TYPE_SCHEMAS",
   "STACK_DEFINITION_COLLECTIONS_SHAPE + COMPOSE_KEY_DISPOSITIONS (picklists, picklistExtensions)",
   "PLURAL_TO_SINGULAR, then regenerated meta-url-data.generated.ts",
   "METADATA_DOCUMENT_TRANSLATORS, so TRANSLATABLE_METADATA_TYPES",
   "data barrel + root export definePicklist",
   "liveness: GOVERNED list, picklist.json (all planned), field.json picklist (planned + authorWarn; the lint warning measured firing), translation.json picklists (planned), README table row + heading 40 to 41, state-counts shards",
   "generated: authorable-surface data/system, json-schema.manifest, api-surface data/root/system + signatures, export-origins, declaration-map, 12 reference docs incl. new data/picklist.mdx",
   "strictness ledger triage row + counts shards; llms.txt data 30 to 31, total 203 to 204",
   "platform-objects metadata-forms bundles: en regenerated; zh-CN/ja-JP/es-ES translated by hand",
   "examples/app-showcase KIND_COVERAGE.picklist + STACK_COLLECTION_COVERAGE.picklistExtensions (waived, linked to #19519)",
   "scripts/check-stack-collection-maps.mjs waiver rows (8 sites)",
   "content/docs: metadata-lifecycle registry count 27 to 28; quick-reference Data 'of 29' to 'of 30'"
  ],
  "count_pins_updated": [
   "spec metadata-type-schemas.test.ts '#4001 — registered-type closure is derived, not tallied > reports the campaign number so a reader never has to count' 25/26 to 26/27",
   "spec metadata-create-seeds.test.ts 'surfaces schema-backed authorable types still missing a seed (no silent cap)' KNOWN_UNSEEDED += picklist",
   "spec metadata-form-zod-reconciliation.test.ts 'field: every top-level key the author may write is offered, or its omission is recorded' omit row for picklist",
   "lint lint-liveness-properties.test.ts 'the SHIPPED field ledger warns on ...' [] to ['picklist']",
   "metadata-protocol protocol.code-only-types.test.ts 'covers every code-only type the registry declares' += picklist, plus a probe (11 generated cases)",
   "metadata-protocol protocol.meta-types-degenerate-derivation.test.ts 'control: `field` still serves N top-level properties' 74 to 75",
   "spec test-typecheck-debt.json: field.test.ts entry deleted (graduated)"
  ],
  "regeneration_turned_green": "metadata-url-spelling.test (7 cases), export-origins.test (3), check-liveness.test (19)"
 },
 "governed_merges_test": "node scripts/pm/check-governed-merges.mjs --branch HEAD at 569b88a475: 'governed-surface predicate: 0 of 64 path(s) hit the register', NOT governed, +1228/-111 (1339 lines, under 5000). Head-repo line: NOT MEASURED by --branch. If open question 2 goes to option A, the PR becomes Tier H.",
 "scope6_sibling_card_draft": "Title: picklist kind: CLI compile/validate/lint accept `*.picklist.ts` / `picklists` / `picklistExtensions` (phase 1 of #18164, Scope 6 of #19518). Points: (1) build/compile ingest of the two collections and `os generate picklist` file naming (packages/cli/src/utils/metadata-file-name.ts parity with the registry filePatterns); (2) `os validate` refuses picklist+options (inherited from FieldSchema) and a field whose `picklist` names no picklist in the stack or its declared dependencies (the loud wrong-reference refusal of ruling 5755653853 item 3); (3) lint R8 field/select-missing-options counts `picklist` as an options source (finding above); (4) `os i18n extract` walks `picklists.NAME.{label, options.VALUE}` (check:i18n-walk-parity, open question 3). Blocked-by: #19518. Dedupe words: picklist, picklistExtensions, os validate, i18n extract, select-missing-options."
}

---

Generated by Claude Code

[objectstack-fleet]

Ruling on the report 5908082075, and round 2 · 2026-09-30T09:23Z

domain:spec seat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1), holder of claim 5905425409. Draft PR #20823 at 569b88a475.

Accepted as delivered: Scope items 1 to 5, as listed below. The census of enumerations and count pins, the six ablations, and the graduation of the test-typecheck debt entry are accepted too.

• the picklist kind: registered, loadOrder 8, a stack collection;
• FieldSchema.picklist, and its refusal when options is also declared or when the type carries no options;
• PicklistServedFieldSchema;
• the picklists.NAME translation face and its fallback, with the field-level entry winning;
• additive picklistExtensions.

Open question 1 ("neither refused"): split to #20827 for the maintainer. Refusing a select or radio that has neither key narrows stored data, and the design of record never costed that. #20827 carries the decision analysis (5908178556) and needs-user-decision. This PR does not wait for it. The card body's Scope 2 and Acceptance now point there.

Open question 2 (check:skill-top-level-keys): A. This PR adds picklists and picklistExtensions to skills/objectstack-platform/SKILL.md, and cuts at least the same number of tokens elsewhere in that file. The cut must be text that is redundant or false, and nothing an author needs. With that change the PR is Tier H: it needs an authorized approval before it goes ready. The NORTH-STAR line and the two records-forms checklist items stay in their own Tier H docs PR, after this lands.

Open question 3 (check:i18n-walk-parity): A. The surface widens to packages/cli/src/utils/i18n-extract.ts (collectExpectedEntries walks picklists.NAME.{label, options.VALUE}) and scripts/fixtures/i18n-walk-parity/every-group.stack.json. Neither side can be green alone. #20825 (the Scope 6 sibling, filed for domain:cli) records that this walk rides this PR.

Open question 4: A. The PR stays Part of #19518, because the Tier H docs PR is still owed after it.

Out-of-scope findings:

• The lint R8 finding (select-missing-options does not count picklist) goes to picklist kind: the os CLI compile / validate / lint path accepts *.picklist.ts, picklists and picklistExtensions, and os validate refuses a picklist that names no picklist (Scope 6 of #19518) #20825.
• The served-shape round trip, resolveObjectFieldLabels, and the runtime enumerations go to picklist metadata kind — runtime: resolve picklist → options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519, as the report names.

Round 2: the two gate fixes above. Then CI goes green, the at-tier review runs with needs:contract-review on the PR and this card, and the maintainer's approval follows.

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report

{
 "issue": 19518,
 "status": "done",
 "branch": "claude/issue-19518-picklist-kind",
 "pr": "https://github.com/objectstack-ai/objectstack/pull/20823",
 "session": "session_01Sfe5YjBLwB9J3y8fvm2xq1",
 "premise_still_valid": true,
 "head": "388dcceb29",
 "summary": "Round 2 per ruling 5908214854, appended onto 569b88a475 with no force-push; origin/main was merged twice (3a85fdf5c2, 12896a0902) via os-regen-merge.sh, each followed by a regeneration commit (45dddd7f2b, 388dcceb29). Gate 1: skills/objectstack-platform/SKILL.md now lists `picklists` and `picklistExtensions` (4049ae31ba), paid for by deleting two pieces of text in the same file, so the file shrinks by 7 bytes and 2 tokens. Gate 2: `collectExpectedEntries` walks `picklists.NAME.{label, options.VALUE}` and an extension's options under the list it extends. It follows the datasets block, adds a `picklist` coverage bucket (`i18n/missing-picklist`), a `parity_tier` member in every-group.stack.json and a 6-test pin; the ablation took the gate red (c9fe5047d4). Round 1 missed six census pins that enumerate registered kinds, and CI went red on them (Test Core shards 1-5, and driver-sql). They are fixed here: 12592f1753 (driver-sql builtin-column-collision classifies `picklist` as presentation) and a1f0668666 (CLI metadata-type-schema-gate GATED_AT, metadata-protocol recovery-doors bootstrap set, platform-objects bare-type and metadata-form echo counts, runtime meta list/item census cells). Also fixed: the picklist docblock no longer opens on a level-1 heading (56d7584e0e; spec file-description corpus 32 back to 31). The changeset now names @objectstack/cli and @objectstack/driver-sql (14b1f68221). CI on 388dcceb29 is GREEN, and the PR stays `Part of #19518` (ruling 4A). No labels were written: the dispatch named none, and a changeset exists.",
 "gate_readings": {
  "check:skill-top-level-keys": "before (569b88a475): exit 1, `[missing] picklists` and `[missing] picklistExtensions` at skills/objectstack-platform/SKILL.md:43. after (388dcceb29): exit 0, 'enumerates all 46 authorable top-level keys of ObjectStackDefinitionSchema (2 declared machine-written and held out: viewItems, runtimeModule)'; self-test 14 cases pass.",
  "check:i18n-walk-parity": "before (569b88a475): exit 1, 'picklists — declared in TranslationDataSchema, produced by no emitter'. after (388dcceb29): exit 0, '11 declared group(s), 9 walked, 2 exempted — every declared group has an extractor face'; self-test 43 cases pass. Ablation (lock-held, trap-restored): anchor `walkPicklists(config, out);` hit 1 then 0, marker 0 then 1, cli rebuilt, ablation-dist-preflight marker present exit 0. Gate under ablation exit 1 ('picklists — ... produced by no emitter'); pin test/i18n-picklist-walk.test.ts 5 failed / 1 passed (the one that stays green asserts an absence: no field-level option key for a picklist-bound field). Restore: ablation-replace 'blob == HEAD (37843a0e963d) and git diff HEAD is empty', rebuilt, preflight --absent exit 0 over 552 dist files, gate exit 0 again."
 },
 "skill_md": {
  "added": "In the Full Configuration Reference enumeration, `picklists`, `picklistExtensions` after `objectExtensions`, next to the object keys they serve (+35 bytes; re-wrapped, 1 line added).",
  "deleted": [
   "CLI Commands: 'covered in **Part 3 — Operations** below' became 'in [Part 3 — Operations](...)'. The words 'below' and 'covered in' and the bold are dropped (31 bytes), because the text is DISTORTED. Part 3 is not below: commit 595621d35c (#19738, 2026-09-23) moved it verbatim to references/operations.md, as line 479 of the same file says. The link target is unchanged.",
   "CLI Commands: 'High-level cheat sheet' became 'Cheat sheet' (11 bytes), because it is REDUNDANT: a cheat sheet is by definition high-level, and the block that follows says what it covers (the bootstrap loop). The join also removed 1 line."
  ],
  "readings": "bytes 23313 to 23306 (-7); tokens (ceil(bytes/4), check-skills-token-ratchet) 5829 to 5827, ceiling 5833, headroom 4 to 6; file lines 489 to 489; skills package lines (all SKILL.md) 4396 to 4396; authored ratcheted total 143408 to 143406. No re-wrap was used to buy lines: the +1 line in the enumeration is paid by the -1 line of a content deletion."
 },
 "maintainer_brief_draft": "## 维护者速读（草稿）\n改了什么：objectstack-platform 技能的顶层键清单加上 picklists、picklistExtensions 两个键。同一文件的 CLI 小节有一句指路文字，删掉其中已失真的“below”和重复的“High-level”。整文件 489 行不变，token 从 5829 降到 5827（上限 5833），技能包总行数 4396 不变。\n为什么改：本 PR 给 stack 新增了这两个顶层键。必需门禁 check:skill-top-level-keys 要求技能页的清单与 schema 一致，页面没列出的键，AI 作者就不会写。删掉的两处：Part 3 在 595621d35c 已拆到 references/operations.md，不在本页下方，所以“below”是错的；“High-level cheat sheet”里的 High-level 与 cheat sheet 意思重复。\n风险与代价（含回滚）：只改文字，链接目标不变，作者需要的内容没有删。回滚办法是 revert 4049ae31ba，但回滚后该门禁会重新变红，除非同时去掉这两个 stack 键。\n席位意见：\n你要做的：本 PR 为 Tier H（skills/** 命中 1 个文件），需要您亲手合并，或给出授权的 APPROVED 审阅后由席位落地。请看 SKILL.md 的两处改动（第 43 行起的清单，第 453 行的 CLI 句子）。",
 "governed_merges": "node scripts/pm/check-governed-merges.mjs --branch HEAD at 388dcceb29: exit 3 = GOVERNED, landing tier H (人合); skills/** x1 (skills/objectstack-platform/SKILL.md); 1 of 75 paths hit the register; +1432 / -132 (1564 changed lines, under the 5000 threshold); merge-base 820d3f4f8e, origin/main 05a7547c9f; head repo NOT MEASURED by --branch.",
 "tests": "All runs used os-verify-lock at head 388dcceb29 with VERDICT command-exit 0, and each exit code was captured before any pipe. Build: turbo build of the cli/runtime/showcase/lint/client dependency closures (60 tasks, exit 0), after spec build + `check:generated --fix`, which reports 'All 15 generated artifacts are up to date'. Tests: spec scripts/file-description.test.ts 111 passed; platform-objects, the 2 echo-decision files + sys-job.global-unique, 82 passed; metadata-protocol whole package 191 files, 2818 passed, 19 skipped; runtime meta-list-projection-parity + 4 registry-reading files 741 passed; cli integration metadata-type-schema-gate 8 passed; cli unit project whole 238 files, 3377 passed (includes i18n-picklist-walk 6/6). Registry-reading census files in packages my diff does not touch, run as a declared narrowing so that a hidden red would surface (CI stops a shard at its first red package): lint 5 files 221 passed; metadata-core 2 files 45 passed; metadata 5 files 173 passed; objectql 8 files 207 passed; rest 2 files 24 passed; client 2 files 27 passed; app-showcase coverage.test 13 passed. Typecheck: @objectstack/spec and @objectstack/cli typecheck exit 0 (test-typecheck: spec 52 files/137 pinned signatures held; cli 3 files/6 held). Before this round's fixes, the whole-package runs at 12592f1753 reproduced CI's red exactly: spec file-description 32 vs 31; platform-objects 5 red; metadata-protocol recovery-doors 1 red; runtime parity census 2 red; cli integration gate 1 red ('expected [ picklist ] to deeply equal []'). Gates: dispatch-gates --commands at 388dcceb29 derives 149 families (147 last round, plus check:driver-conformance and check:route-envelope), all run under the lock. Two are non-zero: check:platform-checklist exit 1 (coverage.json picklist UNCLASSIFIED, since docs/qa is off-limits, and the pre-existing identity-auth twoFactor anchor, unchanged) and check:pm-dispatch-gates exit 124 (killed by its 900s timeout; recorded with its reason). check:dual-build-cjs-loads, NOT MEASURED last round, is now exit 0. `dispatch-gates --ran`: '149 derived famil(ies) accounted for — 148 run, 1 NOT-MEASURED (0 DERIVED from a recorded exit 3, 1 a KILL code with a stated reason)', exit 0. NOT MEASURED: check:pm-dispatch-gates (reason: timeout kill, exit 124; the gate reads scripts/pm tooling this PR does not touch). CI on 388dcceb29, deduplicated by name: 26 names over 35 check-runs, with 24 success and 2 skipped (Console Pin Gate, Packed-tarball smoke (opt-in)). Success: Auto Label, Build Core, Build Docs, Check Changeset, Check Documentation Links, Check PR Size, Dogfood Regression Gate x4, Dogfood Verify CLI, Flag docs affected by code changes, Governed Surface Queue Guard, Lint & Repo Gates, No other open PR may claim the same issue, No other open PR may claim the same single-writer path, Part-of PR must not also close its card, Spec property liveness, Temporal Conformance (live PG + MySQL), Test Core x7, The card this PR closes must claim this branch, Type Check · consumer gates, Type Check · debt ledger, Type Check · source gates, Type Check · workspace, TypeScript Type Check, filter. ci-failure.mjs --pr 20823: 'VERDICT: GREEN — all 35 check-run(s) completed, none failed', exit 0.",
 "mcp_calls": "0",
 "api_writes": "1 relay write through scripts/pm/fleet-write/dispatch.mjs (transport dispatch, as objectstack-fleet[bot]): comment POST /repos/objectstack-ai/objectstack/issues/19518/comments (this os-dev-report). Plus git push of the branch (12592f1753 and then 12592f1753..388dcceb29), which is not a REST write. No PR body edit and no label write this round.",
 "open_questions": [],
 "out_of_scope_findings": [
  "carrier: the seat's docs/qa follow-up · check:platform-checklist still reports coverage.json picklist UNCLASSIFIED (docs/qa/platform-checklist is off-limits to this card). The same run shows the pre-existing identity-auth.json ABSENT SYMBOL plugin-auth auth-plugin.ts#twoFactor, which this branch does not touch · noted, not filed.",
  "carrier: the seat · a PR-body update is owed, since dev does not PATCH. The body predates round 2, so the seat adds the 维护者速读 section (draft in maintainer_brief_draft), the governed-merges line and the round-2 file surface (packages/cli/src/utils/i18n-extract.ts + i18n-coverage.ts, scripts/fixtures/i18n-walk-parity/every-group.stack.json, skills/objectstack-platform/SKILL.md, packages/drivers/driver-sql/src/builtin-column-collision.ts, and census test pins in cli, metadata-protocol, platform-objects and runtime) · noted, not filed."
 ]
}

---

Generated by Claude Code