finding(spec): the object.form options repeater offers an icon input that SelectOptionSchema refuses at publish — offer-vs-door disagreement, the half #13669 did not fix

Found while implementing the objectui#6140 / objectui#6153 spec half (PR #13669). Filed unassigned, not fixed there — it is outside that ruling's scope (the ruling declared rows and options[].description only).

The disagreement

packages/spec/src/data/object.form.ts, the fields repeater's options sub-form, offers five inputs per option:

label, value, color — declared on SelectOptionSchema, fine.
description — was offered-but-refused; DECLARED by PR feat(spec): declare rows (multiline editors) and options[].description as authorable field-metadata keys (objectui#6140 / objectui#6153, spec half) #13669 under the objectui#6153 inherited ruling, so this half is now honest.
icon (helpText "Lucide icon name") — still offered, and still refused: SelectOptionSchema (packages/spec/src/data/field.zod.ts) does not declare icon, and the shape is strict, so a value typed into that form input is an unrecognized_keys refusal at publish.

So the authoring form teaches a key the publish door rejects — the author finds out at the 422. This is the same class the #11410 / #12868 work retired elsewhere (a form offering a choice publish refuses), one key over.

Why it is not obviously "just delete the row"

The non-declaration of icon on SelectOptionSchema is deliberate and recorded: the #5016 per-key measurement (documented in packages/spec/src/ui/action.zod.ts, the action-param option docblock) left icon / disabled out with option C — "None has a reader an action param's option list can reach". But that measurement was about the ACTION-param path; objectui's SelectOptionMetadata interface does declare icon (per the same docblock: label / value / color / icon / disabled / visibleWhen), so whether a FIELD option's icon has a live objectui reader today deserves its own measurement before choosing between:

remove the icon row from object.form.ts (declared = offered converges downward), or
declare icon on SelectOptionSchema (needs a measured reader plus a ruling — an accepted-set expansion, Clause-② class).

PR #13669's pin file (field-rows-option-description.test.ts) pins icon as still-refused, so whichever way this resolves, that pin moves with it deliberately rather than silently.

Refs: PR #13669 · objectui#6140 / objectui#6153 (the family ruling this fell out of) · #5016 (the action-path per-key measurement) · #12868 (the form-face option narrowing precedent).

Generated by Claude Code

Triage: lands in packages/spec/src/data/object.form.ts(+ 可能 field.zod.ts)⇒ domain:spec, pm:queue, type Task, priority:p2。

⚠️ 本卡以「带前提的裁决」入队 —— 三件套,缺一不可

卡列了两个选项并说「选哪个需要先测一个读数」。分诊不把它推进决策箱,理由:方向由在案规则机械决定,只有一个前提待测。

① 裁决(不可重裁):offer-vs-door 不一致按 ADR-0049 enforce-or-remove 处置。零具名 reader ⇒ 走 remove(把 icon 行从 object.form.ts 的 options 子表单撤下),declared = offered 向下收敛。

② 前提(具名、可证伪,dev 必须先验再动手):

FIELD option 的 icon 在 objectui today 没有活的 reader。

⛔ 不得用 #5016 的读数代替这次测量 —— 卡已经点明那次测的是 ACTION-param 路径,而 objectui 的 SelectOptionMetadata 接口确实声明了 icon。两者是不同的面,拿 A 面的零去证 B 面的零,正是本仓反复摔过的那种假零。测法:在 objectui 树上找 field-option 渲染路径对 icon 的实际读点,零命中必须带阳性对照(同一遍里 color 或 label 必须命中),否则零不成立。

③ 显式禁令:前提不成立(测到活 reader)⇒ 停手,报 fork,⛔ 不许硬做,也不许悄悄改走选项 2。选项 2 是接受集扩大 + Clause-② 类(卡自己也这么标),它需要维护者裁决,不是 dev 可以顺手选的另一条路 —— 届时本卡回 needs-user-decision,由分诊席重挂。

随卡带的约束

feat(spec): declare rows (multiline editors) and options[].description as authorable field-metadata keys (objectui#6140 / objectui#6153, spec half) #13669 的 pin 必须同笔移动:field-rows-option-description.test.ts 目前把 icon pin 成 still-refused。走 remove 之后该 pin 的含义变了(不再是「schema 拒绝」而是「表单不再提供」)⇒ 显式改写并说明,⛔ 不许静默删除,也 ⛔ 不许留着让它继续断言一件已经换了主语的事。
⛔ 不碰 description:那半边刚由 PR feat(spec): declare rows (multiline editors) and options[].description as authorable field-metadata keys (objectui#6140 / objectui#6153, spec half) #13669 按 objectui#6153 继承裁决 DECLARED 掉,是已裁面。
触 packages/spec ⇒ 派发侧按条款②取档;但注意:remove 方向不改变接受/拒绝行为(icon 今天就是 422,撤下表单输入后仍然是),⇒ 路径肢命中而内容肢大概率 Clause-②: no —— 认领评论照实申报,由入队闸门判,⛔ 分诊不代判。

Generated by Claude Code

Claim: PM loop round 1 (batch 2)
Session: session_01GDA48PuRFrHyRfdkBz8m21
Branch: claude/issue-13671-option-icon-offer-removal
Worktree: objectstack-issue-13671
Domain: domain:spec
File surface: packages/spec/src/data/object.form.ts (the options sub-form icon row at :159 — NOT the object-level icon row at :21), packages/spec/src/data/field-rows-option-description.test.ts (the icon control pin rewritten with its new meaning), any form-face fixture that enumerates the repeater's inputs; ⛔ SelectOptionSchema in field.zod.ts untouched on the remove route; read-only measurement in the objectui checkout (stop on breach; explain in the report)
Container & model: S/M, mode:subagent, model: opus — node scripts/pm/dispatch-gates.mjs --tier packages/spec/src/data/object.form.ts (tree a39b02a6) answered "no path-derived mandate … Clause ② SUSPECT surface: packages/spec/src/data/object.form.ts ⇢ packages/spec/src/**"; tier judged from content — the ruled remove route changes no accept/reject behaviour (icon is refused today and stays refused); the fork (declare icon) is Clause-② class and is a STOP, not a route
Clause-②: no
Serial constraints cleared: no open PR touches object.form.ts or the pin test (open-PR file lists and titles read at fire); batch-1 in-flight #13802 edits field-value.zod.ts, a different file; same-day churn: #14158 (2026-09-01) last touched both files — dev works from current origin/main; generated content/docs/references/** is a shared product — landing relay order applies; none else.

Decision re-read done to the last comment: no maintainer ruling; triage comment 5475462379 is a premise-bearing ruling (ADR-0049 enforce-or-remove ⇒ remove the icon row; premise = no live objectui reader of a FIELD option's icon, measured with a positive control; premise false ⇒ stop and report, never option 2). Quoted verbatim into the dispatch order. Batch cap raised to 5 per the maintainer's instruction of 2026-09-02 (verbatim 「任务很多，并发保持5」).

Generated by Claude Code

Claim (dev): starting implementation under the PM dispatch for this card.

Session: session_01GDA48PuRFrHyRfdkBz8m21 — subagent mode, so this equals the dispatching PM session ID in comment 5502783304; the branch below is the identity bit.
Branch: claude/issue-13671-option-icon-offer-removal (pushed empty at a39b02a6 before the first edit)
Worktree: objectstack-issue-13671
Assignee field untouched — it belongs to the PM seat.

Route per the triage ruling (comment 5475462379): ADR-0049 enforce-or-remove, premise-first. I measure the FIELD-option icon reader question on the objectui tree with a positive control before any edit; premise false (a live reader found) means stop and report needs_decision, never option 2.

Expected file surface: packages/spec/src/data/object.form.ts (the options sub-form icon row only — not the object-level icon row), packages/spec/src/data/field-rows-option-description.test.ts (the pin rewritten with its new subject), plus any generated form-face artifact the repo tooling regenerates. SelectOptionSchema in field.zod.ts stays untouched, and description is not touched.

Clause-②: no — the remove route changes no accept/reject behaviour (icon is refused today and stays refused).

Generated by Claude Code

os-dev-report

{
  "issue": 13671,
  "status": "done",
  "branch": "claude/issue-13671-option-icon-offer-removal",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/14326",
  "premise_still_valid": true,
  "summary": "PREMISE HOLDS: zero live readers of a FIELD option's `icon` in objectui, measured twice with a live positive control on the same object shape — at the `.objectui-sha` pin d8ec8d6d4 (the console this repo ships) and at objectui origin/main 67dadd602, where packages/fields/src/index.tsx reads option?.label (:1538) and option?.color (:1545,:1546,:1567,:1568) off a SelectOptionMetadata[] and never option?.icon; the whole tree's 204 source-level .icon reads include none in packages/fields, and the single opt.icon read (components/src/custom/config-field-renderer.tsx:204) is the config-panel ConfigField vocabulary, typed React.ReactNode and unreachable from an authored field option. So the ADR-0049 remove leg was taken: the `icon` row is withdrawn from the object.form options repeater (the OBJECT-level icon row at :21 is untouched), SelectOptionSchema and `description` are untouched, and no accept/reject behaviour moves (Clause-②: no). The #13669 pin moved in the same stroke rather than being deleted — its `icon` case is kept and re-commented for the door half, and a new section 3 asserts the offer half plus the general invariant that every input the repeater offers names a key the door accepts. An earlier measurement pass whose positive control MISSED (too-narrow receiver list, only .color) was discarded as a broken pattern, not read as a zero.",
  "tests": "All readings at head f95dc728 (final commit); gate union re-derived at that head. (1) pnpm --filter '@objectstack/spec^...' build -> 'No projects matched the filters' — verified as a REAL empty set, not a typo: @objectstack/spec declares zero workspace deps. (2) pnpm --filter @objectstack/spec build -> VERDICT command-exit 0 (via scripts/pm/os-verify-lock.sh, slot issue-13671-dev). (3) pnpm --filter @objectstack/spec test -> 'Test Files 450 passed (450)' / 'Tests 12072 passed (12072)'. (4) pnpm --filter @objectstack/spec exec vitest run --maxWorkers=2 --reporter=verbose src/data/field-rows-option-description.test.ts src/data/form-delete-behavior-options.test.ts -> 'Test Files 2 passed (2)' / 'Tests 46 passed (46)', with all four new section-3 cases named green in the verbose listing (so the block ran; not a zero-match pass). (5) pnpm --filter @objectstack/spec typecheck -> exit 0, including \"check:test-typecheck: OK — @objectstack/spec's test layer compiles under packages/spec/tsconfig.test.json\". NOT-MEASURED guard closed with tsc --listFiles: object.form.ts is in tsconfig.json's program, the pin test is in tsconfig.test.json's — both confirmed present rather than assumed. (6) REVERSE VERIFICATION (fix committed first): trap 'restore' EXIT INT TERM with absolute paths; mutation leg re-inserted the `icon` row and was PROVEN ON DISK, not by the editor's exit code — injected-marker grep count 1, blob 38249da9 vs HEAD blob 32ffb1f4. Observed direction: the 4 new section-3 cases go RED while SECTION 2's door case stays GREEN — the direct demonstration that the old single-half pin could not see the disagreement. Restore leg via `git checkout HEAD -- ABSOLUTE_PATH` and proven by observation, not exit code: hash back to 32ffb1f4 (equal to the HEAD blob, non-empty), git diff HEAD 0 lines, porcelain empty, residual-marker count 0. No dist/ablation leg applies — the pins import ./object.form as source, so vitest compiles src and no dist marker can go stale under them. (7) pnpm --filter @objectstack/spec check:generated -> '✓ All 15 generated artifacts are up to date' (this is the SECOND run; the first reported check:api-surface stale because dist predated a post-build source edit — rebuilt, then green). Nothing needed regenerating: check:authorable-surface, check:api-surface, check:docs, check:liveness all green as-is, which follows from no schema key having moved. (8) pnpm --filter @objectstack/spec check:objectui-pin-citations -> '✅ 8 asserting objectui pin citation(s) match .objectui-sha (d8ec8d6d4), 15 historical citation(s) recorded' — 13 historical before this branch, so both new read-point records entered the gate's population, written in its HISTORICAL spelling (an unrecognised spelling is silent, so this was checked rather than assumed). (9) Spec source audits, all exit 0: check:llms-txt, check:yaml-examples, check:variant-docs, check:empty-state, check:entry-nameability, check:browser-reachable-entries, check:exported-any, check:dual-source-exports, check:liveness. (10) Repo gates, all exit 0 (each captured as 'cmd > file 2>&1; EXIT=$?' before any pipe): check:nul-bytes, check:empty-changeset, check:adr-0087-registration, check:changeset-gate-self-tests, check-changeset-no-major, check:cross-package-test-inputs, check:test-source-alias, check:doc-authoring, check:published-files, check:merge-driver, docs-audit/check-affected-docs, docs-audit/check-drift-comment, pm/check-half-states. (11) scripts/check-test-completeness.mjs -> exit 3, 'PREREQUISITE NOT MET' (it grades a saved turbo test log only CI produces): recorded as NOT MEASURED, which its own text says is neither a red nor a finding. (12) Gate family: node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands, no paths, run at the final head — 53 commands over 3 paths, byte-identical to the derivation taken before the last two commits, so nothing was added late. (13) DECLARED NARROWING — repo-wide `pnpm lint` not run (CI owns it). Evidence the narrowing is a measurement: population read from eslint's own config, not guessed — `eslint --no-inline-config --format json` over the diff returns 3 results whose only message is the changeset .md reporting 'File ignored because no matching configuration was supplied', so eslint's population for this diff is the 2 TypeScript files; file count 3 read from the --format json output; invariance for untouched files holds because this repo runs one eslint.config.mjs that never enables type-aware linting for ANY file (no parserOptions.project, no typed rules), stated and positively controlled in that config's own header at :328. CI status at report time: in_progress (reported at draft-PR time per the standing clause; no CI waiting).",
  "mcp_calls": "1 — one targeted mcp__github__search_issues for the out-of-scope dedupe, after the REST /search/issues endpoint returned 403 in this container (repo-scoped REST reads AND writes were 200, so issue/comment/PR/label traffic all went through authenticated curl, and the objectui measurement through git)",
  "open_questions": [],
  "out_of_scope_findings": [
    "filed as #14327: the #3786 metadata form/Zod reconciliation (metadata-form-zod-reconciliation.test.ts) walks only ONE level of nested form lists, so the options repeater — and 5 other depth-2 lists, measured across the 17 registered forms — is outside its population entirely; that is the structural reason this card's offer-vs-door disagreement survived a green gate whose own stated rule is that a form offering a key the Zod refuses is always a defect. Unassigned, unlabelled, deliberately not ridden on PR #14326."
  ]
}

Generated by Claude Code

ACCEPT — PR #14326 (head `f95dc728`, 3 files +207/−7, 3 commits) · Clause-② in-seat contract review: PASS (path limb only)

Reviewer of record: domain:spec seat, session_01GDA48PuRFrHyRfdkBz8m21. Verified on GitHub and on the fetched branch, not on the report's self-description.

Checklist: draft ✓ · base main ✓ · first line Fixes #13671 ✓ · files = the declared surface exactly — the options sub-form icon row in object.form.ts (the object-level icon row at :21 kept, and now pinned as the form's only icon input), the #13669 pin rewritten rather than deleted, one patch changeset for @objectstack/spec ✓ · SelectOptionSchema and description untouched ✓ · no content/docs/releases/ ✓ · governed-surface test 0/3 ✓ · check-clause2-carriers --pair 14326: both carriers agree ✓ · trial merge onto current origin/main: CLEAN ✓.

Premise (the ruling's falsifiable half): HOLDS — corroborated independently by this seat on the objectui checkout at both the shipped pin d8ec8d6d and objectui origin/main: option.icon reads in packages/fields/src = 0, with the positive controls hitting in the same pass (option.color 4, option.label 6, both trees). The dev's wider sweep (204 tree-wide .icon reads, none in packages/fields; the single opt.icon is the config-panel ConfigField vocabulary) is consistent. Remove leg taken as ruled; option 2 not touched.

Contract judgment list: ① derived judgments — no accept/reject behaviour moves (icon refused before and after; section 2's door pin kept and re-commented); the offer narrows to label, value, color, description = the schema's authorable keys minus visibleWhen — Clause-② content limb not tripped, matching the claim's Clause-②: no; ② semver — patch on a published package for an authoring-form change ✓; ③ flags — none (open_questions: []).

Pin discipline: the #13669 pin moved deliberately, with the subject change written above the new section 3; reverse verification re-inserting the row turned the four offer-side cases red while the door case stayed green — the exact blindness of the old single-half pin, demonstrated on disk with blob hashes. Full spec suite 450/450 files, 12072 tests; typecheck three legs; check:generated 15/15; check:objectui-pin-citations records the two new read points.

Recorded, not a REWORK ground: lint narrowed with the three-piece evidence (population from eslint's own config, count from --format json, no type-aware linting anywhere); check-test-completeness NOT MEASURED (CI's). Out-of-scope finding filed as #14327 (depth-2 form lists sit outside the #3786 reconciliation's population) — existence verified by this seat in the same stroke.

Tier fuse: get_session.external_metadata.last_served_model = claude-fable-5-1 (read 02:04Z, this review window) vs CONTRACT_REVIEW_TIER = 'claude-fable-5' — the literal mismatch is the rollover staleness on record in #14303; reviewed in-seat on the floor reading recorded there (comment 5503129551); provenance for in-seat clearing = the 2026-08-31 ruling.

Clearing needs:contract-review on both carriers in this stroke (additive removal, read back). Landing: ready + auto-merge once every check on f95dc728 is green; pm:dispatched stays until MERGED and reconciled.

Generated by Claude Code

os-dev-report

{
  "issue": 13671,
  "status": "done",
  "branch": "claude/issue-13671-option-icon-offer-removal",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/14326",
  "head": "4663ba209",
  "premise_still_valid": true,
  "summary": "Patch round only — the ACCEPTED implementation was not touched, re-implemented or widened. One commit (4663ba209) on top of f95dc728, containing ONLY the four regenerated platform-objects locale bundles: packages/platform-objects/src/apps/translations/{en,es-ES,ja-JP,zh-CN}.metadata-forms.generated.ts. NATURE OF THE CHANGE, for relay to sibling PRs touching generated files: pure key REMOVAL, 4 files / 16 deletions / 0 insertions, exactly 4 lines removed per file — the single \"fields.options.icon\" entry with its label and helpText — because the accepted diff withdrew the options sub-form icon row from packages/spec/src/data/object.form.ts and these bundles are extracted from the *.form.ts leaves. No key was added, no translation was rewritten, no other package bundle moved (the other 8 extract-config packages stayed byte-identical and reported in sync), and no source, schema or test file was touched in this round. Merge mode preserved the existing es-ES/ja-JP/zh-CN translations of the surviving keys; only the en bundle is source-rewritten by design. No changeset added: the branch already carries .changeset/object-form-option-icon-offer-withdrawn.md for the user-visible change, and regenerating a generated artifact adds no second user-visible fact. PR remains draft, base main; no labels touched, ready not flipped, no auto-merge. TWO DEVIATIONS TO NOTE, neither affecting the code: (1) the dispatch said ONE issue comment, but my standing os-dev contract makes the machine-readable os-dev-report comment mandatory as the durable copy, so I posted it (5503687075) IN ADDITION to the short status comment (5503689718) — no labels, ready flag or auto-merge were touched, which is what the dispatch`s no-other-writes clause enumerated. (2) Observation, NOT filed and NOT changed: the accepted PR body already contains the phrase \"not fixed here: #14327\". It does not close #14327 — GitHub`s closing-keyword grammar needs the keyword adjacent to the reference (\"fixed #14327\" / \"fixed: #14327\") and the intervening word \"here\" breaks the match — so this is safe as written. I left the accepted prose alone rather than edit a reviewed body, but flagging it since the margin is one word.",
  "tests": "BUILD PREREQUISITE (the gate names its own closure; run under scripts/pm/os-verify-lock.sh): `pnpm exec turbo run build --filter=@objectstack/cli --filter=@objectstack/platform-objects --filter=@objectstack/plugin-{approvals,audit,security,sharing,webhooks} --filter=@objectstack/service-{messaging,realtime,storage}` -> 56/56 tasks, lock VERDICT command-exit 0. Then a full `turbo run build --filter=!@objectstack/docs` -> 71/71 tasks, lock VERDICT command-exit 0, run AFTER the bundle commit so every gate below read a dist rebuilt from the committed tree.\n\nRED REPRODUCED FIRST, at f95dc728 with the prerequisite already cleared (so it is a content verdict, not the unbuilt-CLI shape that gate header warns reads identically): `pnpm check:i18n` -> `  platform-objects               DRIFTED (4)` / `check-i18n-bundles: 1 bundle problem(s)` / `  - platform-objects: 4 bundle(s) drifted from the schema`, EXIT=1 (captured by redirect-then-$?, never through a pipe). All 8 other packages printed `in sync (7 bundle(s))` in the same run.\n\nGREEN ON FINAL HEAD 4663ba209: `pnpm check:i18n` -> `  platform-objects               in sync (11 bundle(s))` and `check-i18n-bundles: OK (9 package(s) - all bundles in sync, no undeclared authoring keys).`, EXIT=0.\n\nFULL DERIVED UNION ON FINAL HEAD 4663ba209: `node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack` with NO paths (it derived the change set itself: 7 paths vs merge base a39b02a6b, three-dot, committed 7 / working tree 0 / untracked 0) -> 54 commands. ALL 54 EXECUTED, each exit code captured by redirect-then-$?. 53 exit 0. The single non-zero is `node scripts/check-test-completeness.mjs` EXIT=3, whose own verdict text reads `PREREQUISITE NOT MET - this gate grades a saved turbo run test log, and no log was named` and states verbatim `the local reading for this gate is NOT MEASURED` / `It is not a red, and there is nothing here to fix` — recorded as NOT MEASURED, identical to the accepted round. Notable individual verdicts, all EXIT=0: check:i18n, check:i18n-stale-fill (the ONE family the four bundle files add to the union vs the original 3-path derivation), check:authorable-surface, check:api-surface, check:liveness, check:type-check-coverage, check:type-check-debt, check:published-files, check:cross-package-test-inputs, check:test-source-alias, check:engine-double-contract, check:dual-build-cjs-loads, check:objectui-changeset, check:doc-authoring, check:merge-driver.\n\nAFFECTED-PACKAGE TESTS ON FINAL HEAD (under the shared lock, VERDICT command-exit 0): `pnpm --filter @objectstack/platform-objects run typecheck` EXIT=0; `pnpm --filter @objectstack/platform-objects run test` -> `Test Files  33 passed (33)` / `Tests  518 passed (518)` EXIT=0; `pnpm --filter @objectstack/spec exec vitest run --maxWorkers=2 src/data/field-rows-option-description.test.ts` -> `Test Files  1 passed (1)` / `Tests  32 passed (32)` EXIT=0 (the accepted round`s section-2 door pin and section-3 offer pin both still pass).\n\nSTALE-TREE WARNING, RESOLVED BY MEASUREMENT NOT ASSUMPTION: the derivation printed `STALE TREE - this answer is derived from a tree at least 14 commit(s) behind origin/main, and 2 file(s) it derives from CHANGED` (scripts/pm/check-skill-line-ratchet.mjs, scripts/pm/dispatch-gates.mjs). I did not rebase or merge. Instead I checked out a detached worktree at origin/main 72adb7fcb and re-derived the SAME 7 paths (the script`s own three-dot answer, not a hand-rolled diff) there: the --commands output is BYTE-IDENTICAL to the branch-tree derivation (`diff` clean). So the staleness does not change what this PR owes.\n\nDEVIATION, for the record: `pnpm check:i18n` was NOT a newly discovered family. Re-deriving the ORIGINAL 3-path change set today still lists it at position 38 of 53 — it was correctly derived in the first round and simply not run. The miss was in the running, not the derivation, and CI caught it.",
  "mcp_calls": "0 - every GitHub read and the PR-body write went through authenticated REST via curl (api.github.com), plus git for branch/diff facts. Zero MCP GitHub tool calls.",
  "open_questions": [],
  "out_of_scope_findings": []
}

Generated by Claude Code

Patch round (CI red on check:i18n): regenerated platform-objects i18n bundles, head 4663ba209. Same session, same branch.

https://claude.ai/code/session_01GDA48PuRFrHyRfdkBz8m21

Generated by Claude Code

Patch round reviewed — PR #14326 head 4663ba20 (one commit on top of the accepted f95dc728).

Verified against the diff, not the report: the commit touches exactly the four packages/platform-objects/src/apps/translations/*.metadata-forms.generated.ts bundles, 16 deletions / 0 insertions (the withdrawn fields.options.icon label + helpText per locale). No source, schema, test or changeset file moved; the bundle regeneration rides the existing @objectstack/spec changeset, as every prior bundle-regen commit on main did.
Red reproduced on f95dc728 (platform-objects DRIFTED (4), exit 1) and green on 4663ba20 (in sync (11 bundle(s)), exit 0); the full derived union (54 commands) ran on the final head with one NOT MEASURED (check-test-completeness, CI-owned), same as the accepted round.
Root cause of the miss, for the record: pnpm check:i18n was in the first round's derived gate set and was not run — a running miss, not a derivation miss. The reconciliation gate that landed in PR feat(pm): reconcile the gate families a dev DERIVED against the ones it RAN #14306 is the mechanical answer; nothing further owed here.
Body note: the phrase "not fixed here: finding(gate): the #3786 metadata form ↔ Zod reconciliation walks only ONE level of nested form lists, so every repeater inside a nested list is unreconciled — 6 paths, and it is why the options icon offer survived #14327" is not a closing reference (the keyword grammar needs the reference directly after the keyword, and "here" sits between); the post-merge inventory diff will confirm finding(gate): the #3786 metadata form ↔ Zod reconciliation walks only ONE level of nested form lists, so every repeater inside a nested list is unreconciled — 6 paths, and it is why the options icon offer survived #14327 stays open.
Pre-flip reads on 4663ba20: governed-surface test 0 of 7; trial merge against origin/main (e854a531) CLEAN; main has not moved on any of this PR's files since its base. The ACCEPT + Clause-② PASS verdict above stands unchanged.

Ready + auto-merge follow once every check on 4663ba20 is green.

Generated by Claude Code

test(spec): the second data/ file group's test titles state each cite…

Navigation Menu

Repository navigation

finding(spec): the object.form options repeater offers an `icon` input that `SelectOptionSchema` refuses at publish — offer-vs-door disagreement, the half #13669 did not fix #13671

The disagreement

Why it is not obviously "just delete the row"

⚠️ 本卡以「带前提的裁决」入队 —— 三件套,缺一不可

随卡带的约束

ACCEPT — PR #14326 (head `f95dc728`, 3 files +207/−7, 3 commits) · Clause-② in-seat contract review: PASS (path limb only)

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(spec): the object.form options repeater offers an icon input that SelectOptionSchema refuses at publish — offer-vs-door disagreement, the half #13669 did not fix #13671

Description

The disagreement

Why it is not obviously "just delete the row"

Activity

os-warren commented on Aug 31, 2026

⚠️ 本卡以「带前提的裁决」入队 —— 三件套,缺一不可

随卡带的约束

os-zhuang commented on Sep 2, 2026

claude commented on Sep 2, 2026

claude commented on Sep 2, 2026

os-zhuang commented on Sep 2, 2026

ACCEPT — PR #14326 (head f95dc728, 3 files +207/−7, 3 commits) · Clause-② in-seat contract review: PASS (path limb only)

claude commented on Sep 2, 2026

claude commented on Sep 2, 2026

claude commented on Sep 2, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(spec): the object.form options repeater offers an `icon` input that `SelectOptionSchema` refuses at publish — offer-vs-door disagreement, the half #13669 did not fix #13671

ACCEPT — PR #14326 (head `f95dc728`, 3 files +207/−7, 3 commits) · Clause-② in-seat contract review: PASS (path limb only)