spec(ai): the blueprint's select options declare their lifecycle kind and a blueprint object names its stageField, so an AI-built status field ships the kinds the object schema already requires

Source

objectstack-ai/cloud#2926 (the AI builder: a status field's options get their lifecycle kinds, so a loan's stage path stops drawing 逾期 as the step after 已还). Its Blocked-by: was #23809, which landed as PR #23846 (adad73f714). cloud's framework pin be076dad28a0 carries it: packages/spec/src/data/field.zod.ts:452 declares kind: 'open' | 'won' | 'lost' on a select option, and object.zod.ts refuseStageFieldWithoutKinds (about line 1970) makes it mandatory on every option of the field an object's stageField names.

The gap, measured at `be076dad28a0`

The AI builder streams the spec's own blueprint schema: SolutionBlueprintStrictSchema, extended in cloud only by plan-view-intent.ts. That schema has no slot for either key.

packages/spec/src/ai/solution-blueprint.zod.ts:88 (BlueprintFieldSchema.options) and :320 (StrictField.options) are { label, value } only. The lenient side strips unknown keys on parse.
BlueprintObjectSchema (:100) and StrictObject (:331) carry nameField but no stageField.

So a model cannot state an option's kind or which field is the object's stage, and apply_blueprint has nothing to write. The kind rule the object schema enforces never meets an AI-built object, because the builder never names a stage field.

Scope

BlueprintFieldSchema.options[] and StrictField.options[] gain kind: the same enum SelectOption.kind declares, reused and not re-spelled. It is optional in the loose shape and nullable in the strict shape, per the strict pattern.
BlueprintObjectSchema and StrictObject gain stageField: the object key it maps onto, snake_case, nullable in the strict shape. The descriptions say what the object schema refuses: every option of the field it names declares kind, on a single-value select / radio with inline options.
The strict/loose key-set parity test (solution-blueprint.test.ts) covers both keys; the published strict JSON schema regenerates.
Clause-②: yes (widening); a minor changeset for @objectstack/spec. The blueprint schema's only producer and consumer is cloud's propose_blueprint / apply_blueprint (as measured for spec(ai): BlueprintViewSchema and StrictView declare filters (ViewFilterRule[]) and sort, so an AI-built view carries its own filter (ruling A on cloud#2932) #24129).

The spec seat settles whether stageField rides this card. Measure first: without it, the object schema's mandatory-kind rule never binds an AI-built object, and that rule is the loud half.

Done when

Both shapes declare both keys, the parity test is green, and the published strict JSON schema carries them.
objectstack-ai/cloud#2926 can start once cloud's pin carries this card's merge. That route (spec first, then the pin) is the precedent of objectstack-ai/cloud#2150 and of spec(ai): BlueprintViewSchema and StrictView declare filters (ViewFilterRule[]) and sort, so an AI-built view carries its own filter (ruling A on cloud#2932) #24129 / PR feat(spec): BlueprintViewSchema and StrictView declare filters and sort, so an AI-built view carries its own filter #24152.

Why not a cloud-side extension first

The bounded exception of ruling A on objectstack-ai/cloud#2932 rested on a p1 defect users saw on staging. Here, nothing changes for a user until objectui's record path reads the option's kind: at objectui's pin ef71a09f9e5f, deriveStages (packages/plugin-detail/src/synth/buildDefaultPageSchema.ts:391) maps { value, label } only. That reader is objectstack-ai/objectui's card, filed beside this one. So a second declaration in cloud would buy nothing a spec-first order does not.

_{Filed by the repo:cloud#2 seat (seat post #24145, session session_01Cbdek8SKSHxddGqTPZXdqX). ⛔ Not a claim. Dedupe: a search of this repository for the blueprint option kind / stageField returned nothing; #24129 (views' filters / sort, closed) is the sibling.}

Triage: p1 · `target:v18` · `area:ai` · `domain:spec` · `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-17T17:59Z

Triage seat (objectstack-wide, seat post #6015). ⛔ Not a claim, ⛔ not a dispatch.

Fresh read on origin/main f963d8e2b3: packages/spec/src/ai/solution-blueprint.zod.ts has no kind on blueprint options and no stageField on a blueprint object (0 hits).
Why p1 + target:v18: the protocol-first rule. It is a spec schema widening (Clause-②: yes, minor). The named producer and consumer is objectstack-ai/cloud#2926, whose route (spec first, then the pin) follows the spec(ai): BlueprintViewSchema and StrictView declare filters (ViewFilterRule[]) and sort, so an AI-built view carries its own filter (ruling A on cloud#2932) #24129 / PR feat(spec): BlueprintViewSchema and StrictView declare filters and sort, so an AI-built view carries its own filter #24152 precedent.
Scope = the card's four items. kind reuses SelectOption.kind's enum, ⛔ not re-spelled. Whether stageField rides this card is the domain:spec seat's call after measuring, as the card says.
The renderer half is objectui#12907, filed beside this card; it is graded in the next objectui round.

Generated by Claude Code

Claim: PM loop round 2 · 2026-10-17T18:12Z · the AI blueprint's select options gain the lifecycle kind (reusing SelectOption.kind's enum) and a blueprint object gains stageField, in both the loose and the strict shape
Session: session_01TyRhWbVbbQpQJTDbW8VahJ (the seat; the dev is an os-dev subagent of this session)
Account: os-project-manager (the seat's linked user as GET /user answers it; always the card's assignee)
Branch: claude/issue-24198-blueprint-option-kind-stage-field
Worktree: objectstack-issue-24198
Domain: domain:spec
Seat: domain:spec#1 (seat post #6017)
File surface (read on origin/main f963d8e2b; line numbers move; stop on a breach and explain it in the report):

packages/spec/src/ai/solution-blueprint.zod.ts: BlueprintFieldSchema.options (about :90), StrictField.options (about :340), BlueprintObjectSchema (about :100/:109) and StrictObject (about :331/:358).
packages/spec/src/ai/solution-blueprint.test.ts (the strict/loose key-set parity test) and the published strict JSON schema / generated artifacts the build regenerates, by their generators only.
One .changeset/24198-*.md (@objectstack/spec, minor).
⛔ Hold-offs: packages/spec/src/data/field.zod.ts and object.zod.ts (read, not edited); cloud and objectui (objectstack-ai/cloud#2926, record path: a derived stage drops its option's declared lifecycle kind, so terminal stages are still guessed from spelling (deriveStages maps value and label only; spec 18 declares kind) objectui#12907).
Seat's scope call on stageField (the card leaves it to this seat): it rides this card, on the premise that ObjectSchema.stageField is the only trigger of refuseStageFieldWithoutKinds (object.zod.ts, about :3160). If the dev measures another trigger, it reports the fork; it does not drop or widen silently.
Container & model: S treated as M (two keys across two parallel shapes, with describe text an AI model reads), mode:subagent, model: TIER_DEFAULT (dispatch-gates --tier: no path-derived mandate); review ceiling CONTRACT_REVIEW_TIER, reason: Clause-②: yes on a published packages/spec/src/** schema (contract-review.md, 档随面定). This seat serves below that tier, so the review runs as an isolated at-tier subagent, ⛔ never in-seat.
Clause-②: yes
Basis for the Clause-② line: two published schemas gain keys, which widens their accept sets.
Responsibility: n/a — not a defect card (a contract widening for a named producer, objectstack-ai/cloud#2926)
Thread-read: 6238179039
Ruling-ref: none — no maintainer ruling on this card. Its scope is triage's grade 6238179039.
Serial constraints cleared:
All 11 open PRs' file lists were read in this act; none touches solution-blueprint.*.
In-flight claims in the lane: spec(migrations): two step-18 upgrade rationale fragments print a tracker id to the author, "(#11509, ruling A-narrow)", in os migrate meta and the upgrade guide #24179 (PR fix(spec): the step-18 upgrade rationale states the element data-binding ruling in words instead of a tracker id #24201, registry.ts.template), [finding] spec(migrations): the published entry page-template-closed-region-unplaced-refused says a stored page with an unplaced region or an unregistered template "renders as before"; objectui#12476 makes it render a refusal #24195 (this seat, a migration entry file), and the other seats' (lint: a locale pack's picklists.<name> translations are never checked — an option key that is not in the list, or a list no package declares, passes os lint --strict and is dropped silently at runtime #23566, metadata-protocol: duplicatePackage renames a token that names both a prefixed package object and one of its fields as the object, so a copied view's column or form field names a field the copy does not have (family close-out) #23561, refactor(plugin-security,platform-objects,spec): retire the catalog seeders, the per-organization catalog machinery and the four catalog objects; Setup creation is an environment write under single and refused under a wall (ADR-0131 D2/D3/D5/D13) #15204). None names this file.
solution-blueprint.zod.ts was last touched today (9f024d8a2, the spec(ai): BlueprintViewSchema and StrictView declare filters (ViewFilterRule[]) and sort, so an AI-built view carries its own filter (ruling A on cloud#2932) #24129 sibling).

os-dev-report

{
  "issue": 24198,
  "status": "done",
  "branch": "claude/issue-24198-blueprint-option-kind-stage-field",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/24213",
  "session": "session_01TyRhWbVbbQpQJTDbW8VahJ — the parent seat's id (this run is its os-dev subagent; the relay derived the same id from the container)",
  "premise_still_valid": true,
  "summary": "Both blueprint option shapes (lenient BlueprintFieldSchema.options[] optional, strict mirror required+nullable) gain `kind`, which is SelectOption.kind's own enum taken by identity (`SelectOptionSchema.shape.kind.unwrap()`, never re-spelled, field.zod.ts not edited). Both object shapes gain `stageField` beside `nameField` (lenient optional SNAKE_CASE, strict strictIdentOrNull). One shared describe sentence, written from refuseStageFieldWithoutKinds, states what the object schema refuses. The seat's stageField premise held: ObjectSchema.stageField is the only trigger of refuseStageFieldWithoutKinds (one definition object.zod.ts:1970, one call :3160; no other stageField declaration in packages/spec/src; the lint/metadata readers only resolve the pointer), so stageField rides the card. H1-H5 measured true: before the change there were 0 stageField hits and no `kind` key (control: nameField at :109/:358); there are 0 code consumers outside packages/spec (5 non-code hits: a changeset, the auto-generated reference page, ADR-0033; control: 70+11 hits inside). Draft PR #24213 opened as objectstack-fleet[bot], assignee os-project-manager, body read back byte-identical (8829/8829).",
  "tests": "At 5645c2cd9 (the merge of origin/main f7af83945): `pnpm --filter @objectstack/spec exec vitest run --project local --maxWorkers=2` gave 691 files, 20974 passed, 1 todo, exit 0. `pnpm --filter @objectstack/spec typecheck` exited 0 (the test-layer program compiles src/**/*.test.ts). `pnpm --filter @objectstack/spec build && check:generated`: all 14 artifacts up to date, exit 0. solution-blueprint.test.ts has 72 tests. New: a lenient carry/optional/identity/vocabulary/snake_case block, an ObjectSchema contract leg (a 1:1 copy parses; dropping one kind is refused at exactly fields.status.options.3.kind, custom), a strict carry/null/required-at-own-path/vocabulary/snake_case block, and an option-level key-parity pin. The object pin now names stageField. Existing strict fixtures gained stageField:null and kind:null, so the leading-digit pin still fails for the digit only. Ablation through scripts/ablation-replace.mjs in wrap mode (anchor x1 to x0, blob changed, restore proven blob==HEAD and git diff HEAD empty; the test reads src, so there is no dist step): strict stageField renamed to stageFieldAblated gave 20 failed / 52 passed (incl. 'the OBJECT schemas carry exactly the same keys'); lenient option kind renamed to kindAblated gave 5 failed / 67 passed (option key parity, carry, identity, vocabulary, ObjectSchema contract). Both went red, the predicted direction.",
  "gates": "Local, at 5645c2cd9: `node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack` derived 108 commands from merge-base f792290df, 5 paths. All were run sequentially, each exit captured before any pipe. `--ran` printed 'Run reconciliation — 108 derived, 107 run, 1 NOT-MEASURED, 0 UNRUN' and '✓ dispatch-gates --ran: 108 derived famil(ies) accounted for', RAN_EXIT=0. NOT MEASURED: check:dual-build-cjs-loads, reason: exit 3 PREREQUISITE NOT MET, it needs a whole-workspace pnpm build (declared narrowing; CI Lint & Repo Gates runs it). Six first readings (check:doc-formula-expressions 3, check:doc-security-posture 3, spec check:generated 1, check:skill-examples 3, check:lean-entry-closure 3, plus the dual-build one) were prerequisite refusals or a dist read mid-rebuild by my own concurrent turbo build of lint. After building lint/formula/objectql, all 13 gates in the overlap window were re-run, each exit 0. CI on PR #24213 head 5645c2cd9 at report time: 'No other open PR may claim the same issue' success, 'Packed-tarball smoke (opt-in)' skipped, every other check-run in_progress (Check Changeset, Governed Surface Queue Guard, Lint · ESLint, Lint · PM dispatch-gates, Lint · docs & code guards, Lint · repo & tooling gates, Lint · CI, release & engine gates, the Type Check lanes, Spec property liveness, The card this PR closes must claim this branch, …). Test Core and Build Core had not been listed yet: in_progress, not waited on.",
  "line_budget": "n/a",
  "files_changed": [
    "packages/spec/src/ai/solution-blueprint.zod.ts",
    "packages/spec/src/ai/solution-blueprint.test.ts",
    "packages/spec/authorable-surface/ai.json (generated: +ai/BlueprintObject:stageField)",
    "content/docs/references/ai/solution-blueprint.mdx (generated: gen:docs)",
    ".changeset/24198-blueprint-option-kind-stage-field.md (@objectstack/spec minor)"
  ],
  "deviations": [
    "Clause-② spelling: the card's body says 'Clause-②: yes (widening)', but the claim carries 'Clause-②: yes'. The PR body and the changeset both carry the claim's line verbatim, 'Clause-②: yes'; no arm was added on my own.",
    "check:dual-build-cjs-loads not measured locally (needs a whole-workspace pnpm build); left to CI.",
    "test:repo not run locally. The derived list did not name it, and the diff edits no repo-project test. scripts/solution-blueprint-header-row.test.ts reads solution-blueprint.zod.ts's header. The header block is unchanged (the new docblocks sit after SNAKE_CASE), and check:skill-refs, which republishes that row, passed. CI Test Core runs it.",
    "My own `turbo run build --filter=@objectstack/lint`, started to clear two prerequisite refusals, rebuilt packages/spec/dist while gates 055-063 ran. The 13 gates in that window were re-run after the build, all exit 0.",
    "The first attempt of ablation leg 2 did nothing: the log redirect named a nonexistent directory, so bash refused before anything ran (git diff HEAD stayed 0 bytes). It was re-run with the correct path, and that run is the one cited.",
    "origin/main moved 6 commits after the pre-PR merge (rest, service-settings, cloud-connection, metadata-protocol, spec api/protocol + liveness). Those paths do not overlap this diff, so no second merge (§10: merge only on a conflict, a main-caused red or a queue removal).",
    "4 git pushes (empty-branch probe, implementation commit, generated-artifacts+changeset commit, merge of origin/main). These are the push leg of the write budget, per the push-each-step rule."
  ],
  "mcp_calls": "0 — no MCP GitHub tool was called",
  "api_writes": "2 relay strokes via scripts/pm/fleet-write/dispatch.mjs (each one POST /repos/objectstack-ai/objectstack/dispatches executed as objectstack-fleet[bot]). Stroke 1, pr_create: POST /repos/objectstack-ai/objectstack/pulls (draft) + POST /repos/objectstack-ai/objectstack/issues/24213/assignees (os-project-manager), run 39062176125 success. Stroke 2: this os-dev-report comment, POST /repos/objectstack-ai/objectstack/issues/24198/comments. No labels were written (no skip-changeset; a changeset ships).",
  "open_questions": [],
  "out_of_scope_findings": [
    "carrier: none (承接者:无) · noted, not filed — the blueprint carries only the string form of stageField. ObjectSchema.stageField also takes `false` (status field declared non-linear, stage heuristics suppressed). The card scopes a snake_case name, and no producer or consumer of the blueprint `false` form is named. Recorded in the PR's Acceptance notes."
  ]
}

Generated by Claude Code

ACCEPT → PR #24213 comment 6240066851 · contract review PASS 6239776296 · ready and auto-merge armed 2026-10-17T20:02Z · domain:spec seat 1 · session session_01TyRhWbVbbQpQJTDbW8VahJ

Generated by Claude Code

Landed · PR #24213 → `6ec0e12228` (`Fixes #24198`, blueprint select options declare their lifecycle `kind` and a blueprint object names its `stageField`) · `domain:spec` seat 1 · `os-project-manager` · session `session_01TyRhWbVbbQpQJTDbW8VahJ` · 2026-10-17T20:37Z

probe	reading
PR object	merged 2026-10-17T20:34Z out of the merge queue (entered 2026-10-17T20:03Z); merge commit `6ec0e12228` on parent `b7f29ddfe`
`origin/main`	`git merge-base --is-ancestor 6ec0e12228 origin/main` → exit 0 (control: the PR head `5645c2cd9f` → exit 1, a squash merge)
merged vs reviewed head	`git diff b7f29ddfe 6ec0e12228` and the reviewed head `5645c2cd9f` over its merge base `f7af83945` share `git patch-id --stable` `c330dc2e` (ACCEPT `6240066851`, contract review PASS `6239776296`)
queue branch	no `gh-readonly-queue/main/pr-24213-*` ref remains
mis-close scan	`Fixes #24198` is the squash message's only closing keyword. Its other references are objectstack-ai/objectui#12907 and objectstack-ai/cloud#2926, and neither is a closing keyword. This card closed `completed` by that PR alone

pm:dispatched is cleared in this act.

Downstream: objectstack-ai/cloud#2926 (apply_blueprint) can start once cloud's pin carries 6ec0e12228. objectstack-ai/objectui#12907 is the renderer half.

Generated by Claude Code

Navigation Menu

Repository navigation

spec(ai): the blueprint's select options declare their lifecycle `kind` and a blueprint object names its `stageField`, so an AI-built status field ships the kinds the object schema already requires #24198

Source

The gap, measured at `be076dad28a0`

Scope

Done when

Why not a cloud-side extension first

Triage: p1 · `target:v18` · `area:ai` · `domain:spec` · `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-17T17:59Z

Landed · PR #24213 → `6ec0e12228` (`Fixes #24198`, blueprint select options declare their lifecycle `kind` and a blueprint object names its `stageField`) · `domain:spec` seat 1 · `os-project-manager` · session `session_01TyRhWbVbbQpQJTDbW8VahJ` · 2026-10-17T20:37Z

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

spec(ai): the blueprint's select options declare their lifecycle kind and a blueprint object names its stageField, so an AI-built status field ships the kinds the object schema already requires #24198

Description

Source

The gap, measured at be076dad28a0

Scope

Done when

Why not a cloud-side extension first

Activity

objectstack-fleet commented on Oct 17, 2026

Triage: p1 · target:v18 · area:ai · domain:spec · pm:queue · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-17T17:59Z

objectstack-fleet commented on Oct 17, 2026

objectstack-fleet commented on Oct 17, 2026

objectstack-fleet commented on Oct 17, 2026

objectstack-fleet commented on Oct 17, 2026

Landed · PR #24213 → 6ec0e12228 (Fixes #24198, blueprint select options declare their lifecycle kind and a blueprint object names its stageField) · domain:spec seat 1 · os-project-manager · session session_01TyRhWbVbbQpQJTDbW8VahJ · 2026-10-17T20:37Z

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

spec(ai): the blueprint's select options declare their lifecycle `kind` and a blueprint object names its `stageField`, so an AI-built status field ships the kinds the object schema already requires #24198

The gap, measured at `be076dad28a0`

Triage: p1 · `target:v18` · `area:ai` · `domain:spec` · `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-17T17:59Z

Landed · PR #24213 → `6ec0e12228` (`Fixes #24198`, blueprint select options declare their lifecycle `kind` and a blueprint object names its `stageField`) · `domain:spec` seat 1 · `os-project-manager` · session `session_01TyRhWbVbbQpQJTDbW8VahJ` · 2026-10-17T20:37Z