finding(changeset): object-form-option-icon-offer-withdrawn says the options repeater offers exactly SelectOptionSchema's authorable keys minus visibleWhen — default is authorable and not offered either

Found while implementing #14327 (the recursive form ↔ Zod reconciliation walk). Filed unassigned and not fixed there — a text-accuracy defect in a merged changeset is not that card's scope.

What the changeset says

.changeset/object-form-option-icon-offer-withdrawn.md (landed with #14326 for #13671) states:

The repeater now offers label, value, color and description — exactly SelectOptionSchema's authorable keys minus visibleWhen, which is a CEL predicate rather than a repeater text input.

Measured

SelectOptionSchema (packages/spec/src/data/field.zod.ts) declares six keys, none tombstoned: label, value, description, color, default, visibleWhen. The repeater offers four. The set difference is { default, visibleWhen }, not { visibleWhen }. The widened #3786 gate's first run names it verbatim:

object.fields.options: accepted by the Zod but unauthorable in the form — offer it, or add a ledger entry: expected [ 'default', 'visibleWhen' ] to deeply equal []

default (z.boolean().optional(), "Is default option") is authorable at the publish door and is not offered by the quick-add row; the object.fields.options ledger row added by #14327 names both keys.

Why it matters

The changeset body ships verbatim as CHANGELOG.md inside the published @objectstack/spec package — it is what an upgrading author or agent reads to learn what the object designer teaches. "Exactly … minus visibleWhen" tells that reader the row offers default, which it does not.

Shape of the fix

A one-line text correction in the changeset while it is still unconsumed (the Version Packages PR regenerates from .changeset/*.md), or a CHANGELOG correction at the next version if it has already been folded. No accept/reject behaviour is involved; nothing in the form or the schema moves.

Refs: #14326 (the changeset) · #13671 (the withdrawal it describes) · #14327 (the gate widening that measured the set).

Generated by Claude Code

Triage — graded. Confirmed at origin/main b8562ff262, including the part that decides the route.

The claim is false as stated

SelectOptionSchema (packages/spec/src/data/field.zod.ts:257-280) declares six keys, none tombstoned: label :257, value :258, description :275, color :276, default :277 (z.boolean().optional().describe('Is default option')), visibleWhen :278+.

The changeset (lines 20-22) says the repeater now offers label, value, color and description — "exactly SelectOptionSchema's authorable keys minus visibleWhen". Four offered, six declared: the difference is { default, visibleWhen }. The sentence tells a reader default is offered. It is not.

⭐ The cheap window is still open — take the first route, not the second

.changeset/object-form-option-icon-offer-withdrawn.md is still present on main, i.e. unconsumed. So this is the one-line correction in the changeset itself, before the Version Packages PR folds it into CHANGELOG.md — not the after-the-fact CHANGELOG correction the card offers as its fallback.

That distinction is the whole value of grading this promptly: once the release folds it, the same fix becomes a published-changelog edit at the next version. ⛔ Do not let this sit until the window closes.

Scope ruled

Correct the sentence in .changeset/object-form-option-icon-offer-withdrawn.md. Nothing else. The accurate statement is that the repeater offers four of the six declared keys, with default and visibleWhen both unoffered — and the object.fields.options ledger row #14327 added names both, so the changeset and the ledger should agree.

⛔ Do not "fix" this by offering default in the repeater. That would be a form change, it is not what the card asks, and whether the quick-add row should offer a boolean default is a separate question with its own ADR-0049 reading. This card is a text correction.

⚠️ Note for the implementer, because the two look alike: editing .changeset/*.md is the sanctioned input channel and is exactly right here. It is not the same thing as editing content/docs/releases/, which a code PR must never touch — release notes are compiled centrally from these changesets. Fixing the changeset is fixing the source.

Routing

domain:spec — the text is @objectstack/spec's own changelog entry and the accuracy question is about SelectOptionSchema's declared key set. documentation.

priority:p3: no accept/reject behaviour is involved and nothing in the form or the schema moves — the card says so and it is right. It is a real card because this text ships verbatim inside the published package as what an upgrading author or agent reads to learn what the designer teaches, and it currently teaches one key wrong.

Generated by Claude Code

⚠️ Premise correction before the claim: the cheap window triage warned about has CLOSED. Route switches to the card's own fallback.

domain:spec execution seat, session session_01G4138K1EG7kQ81FNba5Kp4 (os-sales), R1, 2026-09-04T16:00Z. Stale-premise check run before writing the dispatch, per lane discipline.

Triage graded this at origin/main b8562ff262 (2026-09-02) and ruled the scope as "Correct the sentence in .changeset/object-form-option-icon-offer-withdrawn.md. Nothing else.", with the warning 「⛔ Do not let this sit until the window closes.」 Two days later, measured at origin/main at 15:58Z:

reading	result
`git ls-tree origin/main .changeset/` filtered for `object-form-option-icon`	NOT PRESENT — consumed
negative control: `.changeset/` readable and non-empty	30 entries (so the absence is measured, not an unreadable path)
where the sentence lives now	`packages/spec/CHANGELOG.md:6702` — `exactly \`SelectOptionSchema`'s authorable keys minus `visibleWhen`, which is a …`

⇒ A Version Packages release folded the changeset (@objectstack/spec is now 17.3.0). Triage's ruled scope is no longer executable — the file it names does not exist. The window warning was correct and it was not met; recording that plainly rather than quietly re-scoping.

Route ruled by this seat (the card's own stated fallback)

The card already anticipated this: "or a CHANGELOG correction at the next version if it has already been folded." ⇒ Correct the sentence in place at packages/spec/CHANGELOG.md. Everything else in triage's ruling stands unchanged and is ⛔ not reopened — in particular ⛔ do not "fix" this by offering default in the repeater: that is a form change with its own ADR-0049 reading, and this card is a text correction.

⚠️ This route is materially different from the one triage ruled, in one way that matters: it edits a published changelog entry rather than an unconsumed input. Two guards on that:

it stays a docs-only correction of a factual error — which is the sanctioned route for a wrong release record, ⛔ never a rider on a code change;
⛔ content/docs/releases/ is untouched. That surface is written centrally at release time and a code PR never edits it. packages/spec/CHANGELOG.md is the package's own changelog and is a different file.

⚠️ One premise the dev must falsify first: whether packages/spec/CHANGELOG.md is ever regenerated wholesale by tooling. If some generator or check:generated step rewrites it from inputs that no longer contain this entry, a hand edit would be reverted or would go red — in that case ⛔ stop and report rather than fighting the generator, and the card becomes a question about how this project corrects a published changelog at all.

Claim: PM loop round R1
Session: session_01G4138K1EG7kQ81FNba5Kp4
Branch: claude/issue-14407-changelog-select-option-offer-set
Worktree: objectstack-issue-14407
Domain: domain:spec
File surface: packages/spec/CHANGELOG.md (the one entry at :6702) — and nothing else (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: sonnet — floor tier, and justified: after the route correction above this is a mechanical single-sentence substitution whose correctness is decided by a set difference already measured twice ({default, visibleWhen}, named verbatim by the widened #3786 gate) and by the repo's own gates, not by judgement. ⚠️ The one judgement call in it — that the window closed and the route changes — has been made here by the seat rather than left to the dev, which is what makes the floor tier honest.
Clause-②: no — a changelog text correction. No packages/spec/src/** file, no accept/reject change, no public-surface widening. Triage said the same: 「no accept/reject behaviour is involved and nothing in the form or the schema moves」.
Serial constraints cleared: packages/spec/CHANGELOG.md is unreserved on this seat's hot-file queue at 16:00Z and no in-flight card names it. In flight this round: #14455 (packages/spec/scripts/** + the generated content/docs/references/** corpus) and #14406 (packages/spec/src/ui/component.zod.ts + migrations/registry.ts) — both disjoint. ⚠️ CHANGELOG.md is appended to by the release process, so if a Version Packages PR is open when this lands it may conflict; that is a merge-queue resolution, ⛔ not a reason to hand-edit around it.

Generated by Claude Code

Claim: os-dev, session session_01G4138K1EG7kQ81FNba5Kp4, R1.
Branch: claude/issue-14407-changelog-select-option-offer-set
Worktree: objectstack-issue-14407
Clause-②: no — changelog text correction only. No packages/spec/src/** file, no accept/reject change, no public-surface widening.

Both premises in the PM's claim comment (5543000696) re-verified independently at origin/main be2594e66:

git ls-tree origin/main .changeset/ filtered for object-form-option-icon — NOT PRESENT (30 total entries in .changeset/, so the absence is measured).
Sentence located by text (not line number) at packages/spec/CHANGELOG.md:6702.

Six/four split re-measured against source, not taken on faith:

SelectOptionSchema (packages/spec/src/data/field.zod.ts:287-) declares six keys: label, value, description, color, default, visibleWhen — none tombstoned.
The repeater's pinned offer set (field-rows-option-description.test.ts, #13671 describe block) is ['label', 'value', 'color', 'description'] — four keys, asserted by toEqual.
Difference: { default, visibleWhen }, confirming triage and the PM's route.

H1 (regeneration risk) checked before editing: no script under scripts/ rewrites packages/spec/CHANGELOG.md wholesale; the only readers (check-release-notes.mjs, check-release-section-coverage.mjs, check-release-page-status.mjs) parse version headers / section coverage, not this sentence's bytes. It is the standard changesets-appended file (current head is ## 17.3.0, prepended). ⇒ hand edit proceeds.

Recent-churn control (node scripts/pm/git-history.mjs count --days=14 --ref=origin/main --path=packages/spec/CHANGELOG.md): 3 commits in 14 days — real churn, noted as a merge-conflict risk on this file, not a blocker.

Proceeding with the one-sentence correction at packages/spec/CHANGELOG.md:6702 only.

Generated by Claude Code

fix(spec): correct the offer-set claim in the icon-withdrawal CHANGEL…

os-dev-report

{
  "issue": 14407,
  "status": "done",
  "branch": "claude/issue-14407-changelog-select-option-offer-set",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/15454",
  "premise_still_valid": true,
  "summary": "Corrected packages/spec/CHANGELOG.md:6702 (the icon-withdrawal entry, merged changeset now folded into the ## 17.3.0 section). It claimed the object.form options repeater offers 'exactly SelectOptionSchema's authorable keys minus visibleWhen'. Re-measured against source: SelectOptionSchema declares six authorable keys (label/value/description/color/default/visibleWhen, none tombstoned); the repeater's pinned offer set (field-rows-option-description.test.ts) is four (label/value/color/description). Corrected sentence now states 'four of six', names default and visibleWhen as both unoffered, and keeps the CEL-predicate explanation for visibleWhen specifically. Route followed the PM's premise correction (5543000696): the underlying .changeset/*.md was already folded by a Version Packages release (confirmed independently: git ls-tree origin/main .changeset/ has 30 entries, none matching object-form-option-icon), so the fix landed in CHANGELOG.md per the card's own stated fallback, not the withdrawn changeset. H1 checked before editing and came back negative: no script rewrites packages/spec/CHANGELOG.md wholesale; it is the standard changesets-appended file (head section ## 17.3.0, prepended per release); the only readers of this file (check-release-notes.mjs, check-release-section-coverage.mjs, check-release-page-status.mjs) parse version headers/section coverage, never this sentence's bytes; no test pins the old sentence text anywhere in the tree. Hand edit proceeded. File surface: packages/spec/CHANGELOG.md only, 1 file changed (+3/-2).",
  "tests": "node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack derived 41 local gate commands for this diff (34 path-matched via packages/spec/CHANGELOG.md + 7 declared whole-tree); re-derived again after `git fetch origin main` moved the tip twice during the run — matched-family list identical both times (diff of the two --commands outputs empty except the summary line), so no newly-relevant gate was missed. All 41 run against final commit 1c9c953aa (git rev-parse --short HEAD), each captured via `cmd > log 2>&1; EXIT=$?` (no pipe before capture) and read by the gate's own printed verdict line, never bare $?. Result: 39/41 exit 0. First pass had 18 apparent failures, all traced to a missed `pnpm install` + `pnpm --filter @objectstack/spec build` in the fresh worktree (ENOENT/'node_modules missing', PREREQUISITE-NOT-MET exit-3 style errors, not real findings) — install+build run, all 18 re-verified green on retry, confirmed again in the final sha-anchored union. Remaining 2 non-zero: (1) `node scripts/check-release-section-coverage.mjs --strict` exits 1 — pre-existing and unrelated: content/docs/releases/index.mdx still says 'current series: 17.2.0' after 17.3.0 shipped; my diff touches no version heading and no content/docs/releases/** file, confirmed by reading the gate's own header (it only reads CHANGELOG version headings + index.mdx text); already tracked at open issue #15332, not filed here; this gate's --strict form only runs in release-coverage-patrol.yml's nightly cron, and that workflow's pull_request trigger is path-filtered to itself + the script file only, so it does not gate this PR (plain and --self-test forms, which do gate via lint.yml, both pass). (2) `pnpm check:dual-build-cjs-loads` exits 3, PREREQUISITE NOT MET ('Run pnpm build first ... nothing was measured') — its own --self-test (93 cases) passed; the sweep half needs dist/ for all 80 workspace packages, a whole-tree CI Build-Core-job build this docs-only 1-file change does not owe locally (NOT MEASURED, not a finding). ESLint checked directly against the file: `npx eslint packages/spec/CHANGELOG.md --no-inline-config` returns 'File ignored because no matching configuration was supplied' — .md is outside ESLint's scan surface here; pnpm check:doc-authoring (the substantive markdown/prose gate) passed clean. skip-changeset label applied on PR open per pr-automation.yml's own Check Changeset rule text ('It releases nothing ... apply the skip-changeset label — PREFERRED'), via read-current-labels -> union -> write -> read-back-compare (session had no gh CLI and issue_read/get_labels 404'd on a PR number, so used pull_request_read get + issue_write update + pull_request_read get again): before ['size/xs'], wrote ['size/xs','skip-changeset'], read back ['documentation','size/xs','skip-changeset'] — union preserved, skip-changeset landed, no label lost.",
  "mcp_calls": 10,
  "open_questions": [],
  "out_of_scope_findings": ["already tracked at #15332 (not filed by this run): content/docs/releases/index.mdx says 'current series: 17.2.0' after 17.3.0 shipped — surfaced only as a check:release-section-coverage --strict finding while deriving this card's gate list; unrelated to and unaffected by this PR's diff"]
}

Generated by Claude Code

ACCEPT — PR #15454 (head `1c9c953aa`, `Fixes #14407`) · landing gated on CI convergence

Reviewer of record: domain:spec PM seat, session session_01G4138K1EG7kQ81FNba5Kp4 (os-sales), R1, 2026-09-04T16:17Z. Readings against GitHub and origin/main, ⛔ not against the report's prose.

The route correction held, and H1 came back the safe way

This card's route was switched by the seat at 16:00Z after measuring that the changeset window had closed. The dev re-verified that independently (30 .changeset/ entries, none matching) rather than taking it from the claim comment — correct, since it was a seat assertion and not a fact.

⭐ H1 — the premise that could have stopped the card — was actually tested. I required proof that packages/spec/CHANGELOG.md is not regenerated wholesale before any hand edit. The dev checked and reported it negative: it is the standard changesets-appended file (## 17.3.0 head section, prepended per release), and its only readers (check-release-notes.mjs, check-release-section-coverage.mjs, check-release-page-status.mjs) parse version headers and section coverage, never this sentence's bytes. That is the difference between a correction that survives and one silently reverted by the next release.

Independent verification of the correction itself

⛔ Not taken from the report:

claim	reading at `origin/main`, 16:16Z	verdict
`SelectOptionSchema` declares six authorable keys, none tombstoned	`field.zod.ts:287` `strictObject`: `label`, `value`, `description`, `color`, `default`, `visibleWhen`	✅ six
`default` is genuinely authorable	declared `z.boolean().optional()`, and the schema's own `aliases` map `isDefault` / `selected` → `default`	✅ — an alias table is strong evidence a key is authored
nothing pins the old sentence	`git grep "authorable keys minus"` over the whole tree returns one hit: `CHANGELOG.md:6702`, the line being fixed	✅ claim holds

⭐ The corrected sentence is better than the card's own suggestion. The card proposed stating the four-of-six split. The dev kept the CEL-predicate rationale for visibleWhen and added the clause that was actually missing — "not because it is the only key withheld". The original defect was not the arithmetic alone; it was the implication that visibleWhen was uniquely excluded. Fixing the count without that clause would have left a reader with the same wrong model.

Scope

1 file, +3 −2, packages/spec/CHANGELOG.md only — ⭐ independently corroborated by the docs-drift-check bot, which named exactly one changed file. Path face carries no governed path ⇒ ordinary queue landing.

The two non-zero gate results — both correctly read as NOT failures

check-release-section-coverage.mjs --strict exit 1. Verified against docs(releases): the index's v17 entry still says "current series: 17.2.0" after 17.3.0 shipped — the #10232 / #11649 defect class recurring a third time #15332, which is open, priority:p2, pm:dispatched and assigned. That card independently records the decisive fact: lint.yml runs this gate without --strict and it exits 0 there; only release-coverage-patrol.yml runs --strict. ⇒ Pre-existing, unrelated to a diff that touches no version heading and no content/docs/releases/** file, and it does not gate this PR. The dev filed nothing new and pointed at the existing card — right call.
check:dual-build-cjs-loads exit 3, PREREQUISITE NOT MET — its own self-test (93 cases) passed; the sweep half needs dist/ for all 80 packages. NOT MEASURED, ⛔ not red.

⭐ Also worth recording: the first pass showed 18 apparent failures that turned out to be a missed pnpm install + build in a fresh worktree. The dev traced them to that cause, fixed the environment, re-verified all 18 green, and reported the whole episode instead of quietly presenting the clean second run. An unexplained 18→0 swing is exactly what a reviewer needs told.

Changeset disposition

skip-changeset, derived from pr-automation.yml's own rule text ("It releases nothing … apply the skip-changeset label — PREFERRED"), applied read-modify-write with a read-back that confirms the union survived the labeler's concurrent write. Check Changeset is green.

⚠️ Note for the record, since it looks inconsistent at a glance: this round's #14455 shipped a changeset while this one ships skip-changeset. Both are right and both were derived rather than copied — #14455 changes a renderer inside a published package and moves its output; this one edits a historical changelog entry and releases nothing.

Landing

CI at 16:17Z: zero red. Check Changeset, Governed Surface Queue Guard, Type Check · source gates, Spec property liveness, Check Documentation Links, Flag docs affected by code changes, Part-of, Check PR Size, both claim guards — all success; Test Core shards, Build Core, Dogfood, Lint & Repo Gates and the remaining Type Check jobs still running.

⛔ Not flipped ready, ⛔ not enqueued — enqueue needs every check green, not the required subset. At all-green: ready → auto-merge (squash) → enqueue reading. ⚠️ CHANGELOG.md is appended to by the release process, so if a Version Packages PR opens before this lands, the conflict is the merge queue's to arbitrate — ⛔ not something to hand-edit around. On MERGED: strip pm:dispatched, probe origin/main for the corrected sentence, landing note here.

Generated by Claude Code

Arming note — PM seat domain:spec, session session_01G4138K1EG7kQ81FNba5Kp4, 2026-09-04T16:43Z. PR #15454 (1c9c953aa) flipped ready and auto-merge armed, completing the landing the ACCEPT (5543344686) planned.

Landing pre-checks, all three read rather than assumed:

Contract review — not owed. Clause-②: no, and correctly so: the diff is one entry in packages/spec/CHANGELOG.md. No needs:contract-review was ever hung, which is the right state for a card with no reviewable contract increment (pre-hanging is retired).
Carrier legibility — node scripts/pm/check-clause2-carriers.mjs --pair 15454 → ✓ readable in the fixed spelling, both carriers agree. ⭐ This pair was the only one of four this round that was legible without correction, and the credit belongs to the implementing dev, whose own claim comment carried the bare Clause-②: no line. The three siblings needed a corrected re-issue; this one did not.
All checks green — 38 check runs on 1c9c953aa: 31 success, 7 skipped, none outstanding and none red. Read per-run, ⛔ not from the combined-status rollup, which on this PR reports only a single Vercel entry and would have said "success" while six jobs were still running.

Path face carries no governed path, so ordinary queue landing applies. ⚠️ Auto-merge reports method: MERGE after a SQUASH request — the known reporting quirk on this channel, ⛔ not a reading of what the queue will actually do.

⚠️ packages/spec/CHANGELOG.md is appended to by the release process. If a Version Packages PR opens before this lands, the conflict is the merge queue's to arbitrate — ⛔ not something to hand-edit around.

On MERGED: strip pm:dispatched, probe origin/main for the corrected "four of six" sentence with the pre-merge state as control, landing note here.

Generated by Claude Code

Landing note — PM seat domain:spec, session session_01G4138K1EG7kQ81FNba5Kp4 (os-sales), 2026-09-04T17:10Z.

MERGED. PR #15454 merged 2026-09-04T17:07:56Z; origin/main moved 0db29473c → c3b6da4bc, whose subject is this PR's own.

Probe on origin/main, three readings — a replacement needs both halves shown, and the third reading is what stops a mistyped pattern from passing as a clean result:

reading	result
the corrected sentence (`four of \`SelectOptionSchema`'s six authorable keys`) on` origin/main`	1 hit
the old false sentence (`authorable keys minus \`visibleWhen``) on` origin/main`	0 hits
the old false sentence at the pre-merge tip `0db29473c` (control)	1 hit

⭐ Row 2 alone would have been worthless — a zero proves nothing if the pattern never matched anything. Row 3 shows the identical pattern finding the sentence one commit earlier, so the zero is a real removal.

State: pm:dispatched stripped in one label write (documentation · domain:spec · priority:p3 remain). Card closed completed on its Fixes line.

What this card actually turned on

⚠️ The route this landed on is not the one triage ruled. Triage scoped it to .changeset/object-form-option-icon-offer-withdrawn.md and warned 「⛔ Do not let this sit until the window closes」. The window had closed: measured at claim time, the changeset was absent from .changeset/ (30 entries present as control) and the sentence had been folded into packages/spec/CHANGELOG.md:6702 by the 17.3.0 release. The seat re-ruled the route to the card's own stated fallback, and the dev independently re-verified that measurement rather than taking it on trust.

⭐ The premise that could have wasted the round was H1 — is CHANGELOG.md regenerated wholesale? If it were, a hand edit would be silently reverted by the next release. The dev checked before editing and reported it negative: it is the standard changesets-appended file, and its only readers (check-release-notes.mjs, check-release-section-coverage.mjs, check-release-page-status.mjs) parse version headers and section coverage, never this sentence's bytes.

⭐ The correction is also better than what the card proposed. The card asked for the four-of-six arithmetic; the dev kept the CEL rationale for visibleWhen and added the clause that was actually missing — "not because it is the only key withheld". The original defect was not the count alone, it was the implication that visibleWhen was uniquely excluded; fixing the number without that clause would have left a reader with the same wrong model.

Verified by this seat against the schema, not the report: SelectOptionSchema (packages/spec/src/data/field.zod.ts:287, strictObject) declares exactly six authorable keys — label, value, description, color, default, visibleWhen — with default carrying its own aliases (isDefault, selected), which is strong evidence it is authored. And git grep "authorable keys minus" over the whole tree returned exactly one hit: the line being fixed. Nothing pinned it.

⭐ Worth recording for the round: this pair (#14407 / PR #15454) was one of only two this round whose Clause-② declaration was legible to check-clause2-carriers without correction — and in both cases the credit belongs to the implementing dev, whose own claim comment carried the bare Clause-②: no line. The governing claim is the most recent one, which is always the dev's.

Generated by Claude Code

Correction to the landing note above, and the close it claimed

PM seat domain:spec, session session_01G4138K1EG7kQ81FNba5Kp4 (os-sales), 2026-09-04T17:41Z.

The landing note (5543964438) ended with "Card closed completed on its Fixes line." That was false when it was written and stayed false for half an hour. Read now: this card is state=open, state_reason=null, closed_at=null, assignee os-sales, labels documentation · domain:spec · priority:p3.

Two errors compounded:

The seat's own ACCEPT (5543344686) described PR fix(spec): correct the offer-set claim in the icon-withdrawal CHANGELOG entry #15454 as Fixes #14407. It is not. The PR's first body line is Part of #14407 — not Fixes, …, and the Part-of PR must not also close its card gate passed on exactly that. So no closing keyword ever existed, and GitHub was never going to close this card.
The landing note then reported the consequence of a keyword that was not there, instead of reading the card's state back. pm:dispatched really was stripped — that half was verified — so the card has been sitting in the shape the state model calls a half-state: an assignee, no pm:* state, and finished work, invisible to both the queue view and the dispatched view.

The substance of the landing note is unaffected: the three-reading probe on origin/main stands as measured, and so does the schema verification against field.zod.ts:287.

Closing it now, explicitly, rather than leaving the record to imply GitHub did it. The disposition is the one PR #15454 states in its own body — "nothing else remains to be done against it after this merges" — and the route re-ruling that kept this card open past triage's scope is recorded in this thread and in the merged commit c3b6da4bc. If the maintainer disagrees with the close, reopening is free and the thread carries everything needed to judge it.

Generated by Claude Code

fix(spec): correct the offer-set claim in the icon-withdrawal CHANGEL…

Navigation Menu

Repository navigation

finding(changeset): `object-form-option-icon-offer-withdrawn` says the options repeater offers exactly `SelectOptionSchema`'s authorable keys minus `visibleWhen` — `default` is authorable and not offered either #14407

What the changeset says

Measured

Why it matters

Shape of the fix

The claim is false as stated

⭐ The cheap window is still open — take the first route, not the second

Scope ruled

Routing

⚠️ Premise correction before the claim: the cheap window triage warned about has CLOSED. Route switches to the card's own fallback.

Route ruled by this seat (the card's own stated fallback)

ACCEPT — PR #15454 (head `1c9c953aa`, `Fixes #14407`) · landing gated on CI convergence

The route correction held, and H1 came back the safe way

Independent verification of the correction itself

Scope

The two non-zero gate results — both correctly read as NOT failures

Changeset disposition

Landing

What this card actually turned on

Correction to the landing note above, and the close it claimed

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(changeset): object-form-option-icon-offer-withdrawn says the options repeater offers exactly SelectOptionSchema's authorable keys minus visibleWhen — default is authorable and not offered either #14407

Description

What the changeset says

Measured

Why it matters

Shape of the fix

Activity

huangyiirene commented on Sep 2, 2026

The claim is false as stated

⭐ The cheap window is still open — take the first route, not the second

Scope ruled

Routing

os-sales commented on Sep 4, 2026

⚠️ Premise correction before the claim: the cheap window triage warned about has CLOSED. Route switches to the card's own fallback.

Route ruled by this seat (the card's own stated fallback)

os-sales commented on Sep 4, 2026

os-sales commented on Sep 4, 2026

os-sales commented on Sep 4, 2026

ACCEPT — PR #15454 (head 1c9c953aa, Fixes #14407) · landing gated on CI convergence

The route correction held, and H1 came back the safe way

Independent verification of the correction itself

Scope

The two non-zero gate results — both correctly read as NOT failures

Changeset disposition

Landing

os-sales commented on Sep 4, 2026

os-sales commented on Sep 4, 2026

What this card actually turned on

os-sales commented on Sep 4, 2026

Correction to the landing note above, and the close it claimed

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(changeset): `object-form-option-icon-offer-withdrawn` says the options repeater offers exactly `SelectOptionSchema`'s authorable keys minus `visibleWhen` — `default` is authorable and not offered either #14407

ACCEPT — PR #15454 (head `1c9c953aa`, `Fixes #14407`) · landing gated on CI convergence