types: re-judge the 27 SCHEMA-NODE-bucketed keys of zod-mirror-parity per UNION ARM — a per-key verdict over per-arm facts let DashboardComponentSchema.widgets' concrete gap hide behind a schema-node sibling (census ordered by the #7952 ruling)

Filed by the director seat under the objectui#7952 ruling (decision batch #68, 2026-09-07).

Why

zod-mirror-parity.test.ts's WiderThanDeclared ledger sorts keys into a SCHEMA-NODE bucket (the recursion-breaking unknown on schema-node mirrors — an instrument artifact, per #7759, excluded from that card's disposition lane). #7952 measured that the verdict is per key while the fact is per arm: DashboardComponentSchema.widgets is a two-arm union whose second arm is a schema-node slot (artifact, correctly bucketed) and whose first arm is a concrete BaseSchema.extend({ type: z.enum(...) }) node with a real TypeScript gap (the ruled-legal metric-card node parses green and tsc refuses it). One key, two arms; the artifact in one absorbed the divergence in the other. #7759 sorted 27 keys out of its lane on that verdict.

Scope

For each of the 27 SCHEMA-NODE keys, enumerate the zod declaration's union arms (if any) and classify per arm: schema-node recursion (artifact) vs concrete shape. Record the table on this card.
Any concrete arm whose TypeScript face lacks it is a real divergence: re-file it into 27 keys across 18 pairs have a zod mirror that ACCEPTS a spelling its TypeScript declaration refuses — each needs a narrow-the-mirror / widen-the-declaration ruling #7759's disposition lane (or as its own card if it carries a ruling, as finding(types): DashboardComponentSchema.widgets is a two-arm union in zod and a one-arm array in TypeScript — the ruled-legal metric-card component node parses green and tsc refuses it, and the parity ledger files the key as SCHEMA-NODE #7952 did) — do not fix here.
Change the ledger's classification granularity so a MIXED key cannot be filed as SCHEMA-NODE wholesale: the entry names the arm(s) that are schema-node and the arm(s) that are concrete. Pin with finding(types): DashboardComponentSchema.widgets is a two-arm union in zod and a one-arm array in TypeScript — the ruled-legal metric-card component node parses green and tsc refuses it, and the parity ledger files the key as SCHEMA-NODE #7952's key as the positive control (a concrete arm must not be hidden by a schema-node sibling).

Acceptance

per-arm table for all 27 keys
each concrete-arm divergence re-filed with a link
ledger granularity changed and pinned

Refs #7952 (the instance and ruling), #7759 (the lane), objectstack#8593 (the ruling that admits the component-node arm).

Claim: PM loop round R2
Session: session_01QtGhnU3WnnWyiWeYQhw2aX
Branch: claude/issue-8252-parity-ledger-per-arm
Worktree: objectui-issue-8252
Domain: domain:spec
File surface: packages/types/src/__tests__/zod-mirror-parity.test.ts (the WiderThanDeclared ledger and its classification granularity) — ⭐ this file is THIS card's for the round; plus the per-arm table posted to this card, and any re-filed divergence cards. ⛔ NOT in surface: packages/types/src/** declarations and mirrors — this card classifies and re-files, it does not fix (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: claude-opus-5 (TIER_DEFAULT — proactive downgrade under the maintainer's standing instruction 2026-09-07 「fable不多了，后續推 opus 任务」; clause ② not met, this changes a test ledger's granularity and touches no published accept set or public surface)
Clause-②: no
Serial constraints cleared: ⚠️ zod-mirror-parity.test.ts is the lane's hottest file and three queued cards read it — #7759 (the disposition lane whose 27 keys this card re-judges) and #7760 (the recursion annotation) are ⛔ NOT dispatched and are held behind this card: their defect shapes and fixes differ from this one, so the fold-or-serial five gates fail and the answer is serial, chain head here. · #7105 IS dispatched in this same batch and is coupled to this file semantically, not textually — see below · #6939's groups 2–6 remain queued, none in flight · #7952 (this card's parent ruling) MERGED as c842594, so this card is unblocked

The coupling with #7105, declared rather than discovered

#7105 is dispatched in this batch and adds action to both faces of EmptySchema. This ledger registers that pair — zod-mirror-parity.test.ts:693 'feedback.zod.ts#EmptySchema', with both faces imported at :299 and :315. Measured on origin/main c842594.

⇒ If both of #7105's faces move together the pair records zero drift and its dev never opens this file. Its brief says so explicitly and tells it to stop and report if it finds it needs a ledger edit, because this file is yours this round. ⛔ Do not coordinate with it directly; if you see a conflict, it goes to the merge queue.

⚠️ This declaration exists because the previous round produced a near-miss of exactly this shape: two PRs both added a line to packages/types/src/index.ts, a file neither claim named, because in both cases it was a post-dispatch increment. The queue merged them cleanly, but the surface was undeclared. ⇒ If your work needs a file this claim does not name, say so in the report before you land, ⛔ do not just add it.

Adjudication — this card IS the ruling's second half

Filed by the director seat under objectui#7952's ruling (decision batch #68, 2026-09-07, maintainer 「同意」). The card body is the complete specification and its three acceptance boxes are the deliverable. ⛔ Not re-openable:

the SCHEMA-NODE bucket is a per-KEY verdict over a per-ARM fact, so a concrete gap can hide behind a schema-node sibling.

#7952 is the measured instance — DashboardComponentSchema.widgets, whose concrete first arm hid behind a schema-node second arm — and it landed as c842594. ⚠️ Read that PR's diff and its two review comments before you start: they establish what a "concrete arm" looks like when it is real, which is the judgement you are about to make 27 times.

PM mechanism assumptions — measure these, and you are encouraged to falsify them

27 is the card's figure, not a verified one. It comes from 27 keys across 18 pairs have a zod mirror that ACCEPTS a spelling its TypeScript declaration refuses — each needs a narrow-the-mirror / widen-the-declaration ruling #7759's exclusion set. Re-derive the SCHEMA-NODE-bucketed key list from the ledger itself before you classify anything; if you get a different count, yours wins and you say so.
⚠️ A key with no union arms still needs a verdict. The card's framing is per-arm, but a plain schema-node mirror with a single arm is the ordinary case and must land somewhere in the new granularity rather than falling through it.
The re-filing half is a judgement, not a transcription. The card says a concrete arm whose TypeScript face lacks it is "a real divergence: re-file it into 27 keys across 18 pairs have a zod mirror that ACCEPTS a spelling its TypeScript declaration refuses — each needs a narrow-the-mirror / widen-the-declaration ruling #7759's lane, or as its own card if it carries a ruling, as finding(types): DashboardComponentSchema.widgets is a two-arm union in zod and a one-arm array in TypeScript — the ruled-legal metric-card component node parses green and tsc refuses it, and the parity ledger files the key as SCHEMA-NODE #7952 did." ⇒ You decide per finding which of those two it is. Filing everything as its own card floods the queue; filing everything into 27 keys across 18 pairs have a zod mirror that ACCEPTS a spelling its TypeScript declaration refuses — each needs a narrow-the-mirror / widen-the-declaration ruling #7759 buries anything that needs a maintainer. Say which you chose and why, per row.
⛔ You are not fixing any divergence you find. The card is explicit. A concrete-arm gap gets a link, not a patch.

Gates

⚠️ scripts/pm/dispatch-gates.mjs cannot derive for this repo — it refuses rather than guessing ("asked for 'objectstack-ai/objectui', but this checkout is 'objectstack-ai/objectstack' … a sister repo's list is hand-derived"). The list below is hand-derived from objectui's own package.json; treat it as a lead and re-derive against your actual diff:

pnpm --filter @object-ui/types type-check (⭐ the ledger's assertions are compile-time — this is the leg that evaluates them) · lint · pnpm exec vitest run packages/types/ · check:control-bytes · changeset-presence (a test-ledger-only diff may legitimately need none — check what the gate actually says rather than assuming either way).

Capture each exit code to disk before reading any output.

Route

Draft PR, Fixes #8252 on the first line, target main. Merge origin/main before opening. ⛔ Do not flip ready, ⛔ do not enqueue, ⛔ do not enable auto-merge — the seat reviews first. ⛔ Do not leave a second claim: this comment is the identity; check the newest Claim: names your branch and stop if it does not. The assignee field is set and ⛔ not yours to change.

Final report: machine-checkable fields only — gates / line_budget / deviations / files_changed — plus your re-derived key count and the per-arm table. Post it here with the os-dev-report marker, then return it.

Generated with Claude Code

https://claude.ai/code/session_01QtGhnU3WnnWyiWeYQhw2aX

Generated by Claude Code

os-dev-report

{
  "issue": 8252,
  "status": "done",
  "branch": "claude/issue-8252-parity-ledger-per-arm",
  "pr": "https://github.com/objectstack-ai/objectui/pull/8323",
  "head": "afcd389",
  "premise_still_valid": true,
  "premise_note": "The card's SCOPE holds and its three acceptance boxes are delivered. One figure inside it is falsified: see key_count_rederived. The re-filing box is satisfied with ZERO re-filings, which is a finding, not an omission.",

  "key_count_rederived": {
    "card_says": "27 SCHEMA-NODE keys (inherited from #7759's exclusion set)",
    "measured": 25,
    "measured_breakdown": "24 keys whose every arm is the artifact + 1 MIXED key (#7952's own). Ledger total 34 entries / 52 keys / 61 arms; CONCRETE half 27, unmoved since #7759.",
    "two_independent_derivations_agree": "the ledger's own docblock classes (25), and a runtime walk marking every key whose slot reaches a recursion-breaking lazy node (25)",
    "why_it_rotted": "WiderThanDeclared is the only one of the five ledgers with NO figure in this file's header. objectui#7733 pinned three ledgers' entry counts and objectui#8222 their key totals; both name the ledgers they reach and this one is in neither. #7759 published 27 keys / 21 pairs, #8252's card body inherited it, and it went two out with every pin in this file green. Closed here: six figures for this ledger are now in the header, read off WIDER_ARMS and the mirrors.",
    "verdict": "mine wins and I say so"
  },

  "per_arm_table": {
    "union_keys_at_the_slot": 9,
    "single_arm_keys": 43,
    "rows": [
      { "key": "complex.zod.ts#DashboardComponentSchema::widgets", "arms": 2, "verdicts": ["CONCRETE", "SCHEMA-NODE"], "ts_face_lacked_an_arm": "IT DID", "disposition": "#7952 — ruled under objectstack#8593, declared by PR #8296 as c842594. Not re-filed: already ruled and already repaired. This is the positive control." },
      { "key": "app.zod.ts#AppComponentSchema::label", "arms": 2, "verdicts": ["CONCRETE", "CONCRETE"], "ts_face_lacks_an_arm": "yes — the locale-map arm", "disposition": "not re-filed: already in #7759's lane, group A (INLINE-LOCALE)" },
      { "key": "objectql.zod.ts#ObjectGridSchema::label", "arms": 2, "verdicts": ["CONCRETE", "CONCRETE"], "ts_face_lacks_an_arm": "yes — the locale-map arm", "disposition": "not re-filed: already in #7759's lane, group A" },
      { "key": "objectql.zod.ts#ObjectGridSchema::description", "arms": 2, "verdicts": ["CONCRETE", "CONCRETE"], "ts_face_lacks_an_arm": "yes — the locale-map arm", "disposition": "not re-filed: already in #7759's lane, group A" },
      { "key": "form.zod.ts#SliderSchema::defaultValue", "arms": 2, "verdicts": ["CONCRETE", "CONCRETE"], "ts_face_lacks_an_arm": "yes — the scalar arm", "disposition": "not re-filed: already in #7759's lane, group B" },
      { "key": "form.zod.ts#SliderSchema::value", "arms": 2, "verdicts": ["CONCRETE", "CONCRETE"], "ts_face_lacks_an_arm": "yes — the scalar arm", "disposition": "not re-filed: already in #7759's lane, group B" },
      { "key": "form.zod.ts#CalendarSchema::defaultValue", "arms": 2, "verdicts": ["CONCRETE", "CONCRETE"], "ts_face_lacks_an_arm": "yes (disjoint)", "disposition": "not re-filed: already in #7759's lane, group D" },
      { "key": "form.zod.ts#CalendarSchema::value", "arms": 2, "verdicts": ["CONCRETE", "CONCRETE"], "ts_face_lacks_an_arm": "yes (disjoint)", "disposition": "not re-filed: already in #7759's lane, group D" },
      { "key": "layout.zod.ts#ContainerSchema::maxWidth", "arms": 2, "verdicts": ["CONCRETE", "CONCRETE"], "ts_face_lacks_an_arm": "yes — the boolean arm", "disposition": "not re-filed: already in #7759's lane, group C" },
      { "key": "the 24 single-arm SCHEMA-NODE keys", "arms": 1, "verdicts": ["SCHEMA-NODE"], "ts_face_lacks_an_arm": "n/a — one arm, so no sibling to hide behind", "disposition": "artifact; #7759's exclusion, unchanged. Each is a one-verdict row, so it lands in the new granularity rather than falling through the per-arm framing." },
      { "key": "the 19 single-arm CONCRETE keys", "arms": 1, "verdicts": ["CONCRETE"], "ts_face_lacks_an_arm": "per #7759", "disposition": "already in #7759's lane (INLINE-LOCALE / FUNCTION-SLOT / structural)" }
    ],
    "conclusion": "The #7952 shape DOES NOT RECUR. Of the 9 union keys exactly one is in the SCHEMA-NODE bucket and it is #7952's own, already ruled and repaired. The other 8 were each already CONCRETE in #7759's lane, arm by arm. One level down was measured too — every arm's own shape properties, for a union mixing a schema-node arm with a concrete one — and found none.",
    "refilings": 0,
    "refiling_judgement": "Per row, both options were weighed. Neither option applies to any row: filing a card for a divergence already sitting in an open lane is queue noise, and filing into #7759 gains it nothing it does not already hold. Filing nothing is the judgement, not a skipped step.",
    "bound_stated": "The census does NOT answer whether a single-arm SCHEMA-NODE key's WIDER reading has a second, concrete cause UNDERNEATH the artifact. That is a different mechanism from hiding-behind-a-sibling and needs a two-face comparison per key, not an arm enumeration. Recorded in the ledger docblock."
  },

  "gates": {
    "types:type-check": "exit 0 @afcd389 (tsc --noEmit + tsconfig.examples.json + tsconfig.test.json — the leg that evaluates the compile-time assertions)",
    "vitest:packages/types": "140 files / 2687 tests, exit 0 @afcd389",
    "check:control-bytes": "exit 0 @afcd389 (6618 tracked text files)",
    "check-changeset-presence": "exit 0 @afcd389 — 1 changeset, EMPTY frontmatter (the gate's own named exemption for a diff confined to src/__tests__/)",
    "check:lint-coverage": "exit 0 — 46/46 packages, 0 errors",
    "check-governed-queue-guard --test": "NOT GOVERNED — 2 paths against 5 governed surfaces, none matched",
    "lint": "NARROWED, and the narrowing is measured: eslint --no-inline-config --format json on the 1 changed file = 1 file, 0 errors, 0 warnings. (1) population read from eslint.config.js itself (files: **/*.{ts,tsx}); (2) file count = length of the --format json output; (3) invariance: project:/projectService appear ZERO times in eslint.config.js so type-aware linting is off, and no rule in eslint-rules/ reads another file => this diff cannot move any untouched file's verdict. Repo-wide pnpm lint is CI's run.",
    "ci": "in_progress — reported at draft-PR time, not awaited"
  },

  "ablation": {
    "discipline": "each leg committed first, mutated on disk with a before/after grep proving the edit landed, run, then restored. Script carries trap ... EXIT INT TERM with an absolute REPO_ROOT path; restore proven by blob hash 612c8634 equal to HEAD plus an empty git diff HEAD. Run under the shared verify lock (VERDICT command-exit 0), on final head afcd389.",
    "A_collapse_7952_key_to_one_verdict": "on-disk two-arm 1->0, collapsed 0->1; 3 failed / 28 passed — arity, positive control, header figures",
    "B_delete_a_single_arm_row": "on-disk row 1->0; 2 failed / 29 passed — coverage, header figures",
    "C_hand_edit_a_header_figure_24_to_27": "on-disk 1->1; 1 failed / 30 passed — header figures",
    "D_break_the_array_unwrapper": "on-disk 1->0; 2 failed / 29 passed — arity, header figures",
    "honest_note": "Leg D did NOT redden the non-vacuity leg, and that is correct rather than a gap: several keys spell the union directly under optional, so a multi-arm key is still seen. The arity pin is what catches it. Recorded because a non-vacuity leg that looks stronger than it is, is the failure this file exists to prevent."
  },

  "files_changed": [
    "packages/types/src/__tests__/zod-mirror-parity.test.ts (+333/-3)",
    ".changeset/wider-ledger-per-arm-8252.md (+4, empty frontmatter)"
  ],

  "line_budget": "N/A — the skills/** net-line ratchet does not reach this diff (no skills/** path touched). Actual delta: +334/-3 across 2 files, all inside packages/types/src/__tests__/ plus one changeset.",

  "deviations": [
    "DECLARED BEFORE LANDING, per the claim's near-miss warning: my work at one point needed packages/types/src/__tests__/handler-keys-string-any-mirrors-7344.test.ts, a file this claim does not name. The first row-key spelling joined pair and key with a DOT, which put a property-read of one mirror's handler-array key into a file that already names that mirror — anchor A of that census — so it read this file as a THIRD reader of a key whose reader set it pins at one file, and failed, correctly. I did NOT add the file. The separator moved to '::' instead, so the diff stays at one test file plus one changeset and the other census keeps its edge; the reason is recorded beside the constant, including an instruction not to spell the colliding key literally there (the first version of the explanation reproduced the very match it described).",
    "The dispatch says to read PR #8296's 'two review comments'. PR #8296 has ZERO review comments and ZERO reviews; its only PR comments are two bundle-size bot posts. The two Clause-2 contract reviews live on CARD #7952 as comments 5568304348 (FAIL) and 5569569634 (PASS). I read those. No impact on the work; recorded so the pointer is corrected.",
    "Verify lock: objectui has no scripts/pm/, so the shared lock was driven through the objectstack checkout's copy at /home/user/objectstack/scripts/pm/os-verify-lock.sh with OS_VERIFY_LOCK_SLOT=dev-8252-arms. The lock is container-wide, not repo-scoped. The earlier iterative type-check and vitest runs during authoring were NOT taken through it; the ablation and the final measurements were.",
    "origin/main was merged before opening the PR (route requirement) and it brought in #7735, which edits the mirrors these new pins read. Everything was re-measured on the merged head afcd389. Incidental live proof: had a mirror's arm count moved, the header-figures pin reads the mirrors and would have reddened.",
    "No skip-changeset label applied: in this repo that label is read by nothing, and the empty-frontmatter changeset is the gate's own named exemption.",
    "No second claim comment posted; assignee not touched; PR left as draft, not enqueued, no auto-merge."
  ],

  "acceptance": {
    "per_arm_table_for_all_keys": "done — for all 52 ledger keys / 61 arms, not just the SCHEMA-NODE bucket. The bucket is 25, not 27.",
    "each_concrete_arm_divergence_refiled_with_a_link": "done with ZERO re-filings, per row and with the reason stated. The one that WAS hidden is #7952's, already ruled and repaired; the other eight union keys are each already in #7759's open lane.",
    "ledger_granularity_changed_and_pinned": "done — WIDER_ARMS holds one verdict per MEASURED arm; four pins (non-vacuity both directions, coverage both ways, arity against the mirrors, and #7952's key as the named positive control) plus the header-figures pin. All five proved able to fail."
  },

  "open_questions": [],

  "out_of_scope_findings": [
    "noted, not filed: a single-arm SCHEMA-NODE key could in principle carry a second, concrete cause underneath the artifact — a different mechanism from #7952's hiding-behind-a-sibling. Answering it needs a two-face MirrorInputOf/DeclaredTypeOf comparison for each of the 24, not an arm enumeration. Not a reproducible defect and not a contract breach, so it is a bound recorded in the ledger docblock rather than a card.",
    "noted, not filed: #7952's prose says the component arm has 'no schema-node recursion in it at all'. Measured, it reaches SchemaNodeSchema through BaseSchema's own body/children slots. The card's conclusion survives the correction and the corrected fact is now recorded in this ledger; the closed card is not worth reopening for it.",
    "noted, not filed: #7759's body says its FUNCTION-SLOT group is '8 keys' and then tables 5. Its group totals sum to 27 only when E counts as 5, which is the figure its headline uses. An internal inconsistency in an open card's prose, not a defect in the tree."
  ]
}

Generated by Claude Code

ACCEPT — PR #8323 at head `afcd389`, `domain:spec @ objectui` seat, R2, 2026-09-07T12:40Z

Reviewed against GitHub and the tree, not against the report's self-description.

Checklist conclusions

item	reading
PR shape	draft, base `main`, first line `Fixes #8252`; no closing keyword beside another open issue
Head pinning	PR `head.sha` = `afcd389529…`; the report's final head reads `afcd389`; `git ls-remote` on the branch agrees. Three-way match
CI	32 check runs, all `completed`: 29 `success` + 3 `skipped`. Zero failures. `Lint` → success 12:26:26Z, `Type Check` → success 12:28:34Z (the leg that evaluates this file's compile-time assertions); all four test shards green
Scope	measured with `git diff --name-status origin/main...afcd389`: exactly 2 files — `packages/types/src/__tests__/zod-mirror-parity.test.ts` (M) and `.changeset/wider-ledger-per-arm-8252.md` (A), +334/-3. Matches the claim's declared surface with nothing added
`content/docs/releases/`	not touched (checked, not assumed)
Changeset	empty frontmatter with an accurate body — the presence gate's own named exemption for a diff confined to `src/__tests__/`. ⛔ Correct here; this repo has no `skip-changeset` label and an empty changeset is the right form
Path surface	not governed. The dev ran `check-governed-queue-guard --test` itself: 2 paths against 5 governed surfaces, 0 matched ⇒ ordinary landing path, no tier gate owed, no carrier to clear

⭐ The finding inside the delivery — a stale figure that every green pin was blind to

The card asked for 27 SCHEMA-NODE keys. The measurement is 25, re-derived two independent ways that agree (the ledger's own docblock classes, and a runtime walk marking every key whose slot reaches a recursion-breaking lazy node).

The root cause is the part worth keeping: WiderThanDeclared is the only one of the five ledgers in this file with no figure in the header at all. #7733 pinned three ledgers' entry counts and #8222 their key totals — and neither reaches this one. So #7759 published "27 keys / 21 pairs", the director seat's card body inherited it, and the number went two out while every pin in that file stayed green. ⇒ A count with no pin is not a fact, it is a rumour with a citation. The PR closes it by putting six figures in the header, read off WIDER_ARMS and the mirrors, and pinning them.

⭐ Zero re-filings is the finding, not a skipped acceptance box

The per-arm table covers all 52 ledger keys / 61 arms, not just the bucket. Of the 9 union keys, exactly one sits in the SCHEMA-NODE bucket — #7952's own, already ruled under objectstack#8593 and already repaired by PR #8296 (c842594). The other 8 were each already CONCRETE in #7759's open lane, arm by arm.

⇒ The #7952 shape does not recur. The judgement was made per row with both options weighed, and filing nothing was the answer: a card for a divergence already sitting in an open lane is queue noise, and re-filing into #7759 gains it nothing it does not already hold. The census also states its own bound — it does not answer whether a single-arm SCHEMA-NODE key hides a second, concrete cause underneath the artifact, which is a different mechanism needing a two-face comparison rather than an arm enumeration. That bound is recorded in the ledger docblock rather than left implicit.

⭐ The declared collision — handled the way the round asked for

The claim warned that if the work needed a file the claim does not name, it must be said before landing, not silently added. It happened: the first row-key spelling joined pair and key with a dot, which put a property-read of one mirror's handler-array key inside a file that already names that mirror — tripping handler-keys-string-any-mirrors-7344.test.ts's reader anchor, which pins that key's reader set at one file. It failed, correctly.

The resolution moved this PR's separator to :: rather than widening the neighbour's allow-list. ⇒ The neighbouring census keeps its edge and this diff stays at the two files the claim named. That is the right trade, and it is exactly the discipline added after the previous round's undeclared-barrel near-miss.

⚠️ My dispatch brief contained an error, and the dev caught it

I wrote "read PR #8296's two review comments". PR #8296 has zero reviews and zero review comments — its only PR comments are two bundle-size bot posts. The two Clause-② reviews live on card #7952 as comments 5568304348 (FAIL) and 5569569634 (PASS). The dev read the right ones and recorded the correction in its report rather than working from a bad pointer. My error, no impact on the work, and the record is right because the dev fixed it rather than following it.

Ablation and other evidence

Four legs, each committed first, mutated with a before/after grep proving the edit landed, restored by blob hash (612c8634 equal to HEAD) plus an empty git diff HEAD, under a trap with absolute paths: collapsing #7952's key to one verdict → 3 red; deleting a single-arm row → 2 red; hand-editing a header figure → 1 red; breaking the array unwrapper → 2 red.

⭐ Its honest note is better than a clean result would have been: leg D did not redden the non-vacuity leg, and the report says so and explains why that is correct rather than a gap (several keys spell the union directly under optional, so a multi-arm key is still seen; the arity pin is what catches it). A non-vacuity leg that looks stronger than it is, is precisely the failure this file exists to prevent.

Out-of-scope findings, verified as noted-not-filed

Three, all correctly left as notes: the single-arm "second cause underneath" bound (recorded in the docblock); a correction to #7952's prose (its component arm does reach SchemaNodeSchema through BaseSchema's own slots — the card's conclusion survives it, and the corrected fact now lives in this ledger); and an internal inconsistency in #7759's prose (its FUNCTION-SLOT group says 8 and tables 5). None is a reproducible defect or a contract breach. ⛔ No cards owed.

Verdict: ACCEPT. No carrier to clear — flipping ready and enqueueing.

Generated with Claude Code

https://claude.ai/code/session_01QtGhnU3WnnWyiWeYQhw2aX

Generated by Claude Code

MERGED — PR #8323 landed as `645ecb9` on `main`, 2026-09-07T12:57Z

Verified on the tree, not on the merge notification. pm:dispatched stripped and the assignee cleared in the same stroke as this note.

probe	reading on `origin/main` `645ecb9`
`WIDER_ARMS` — the per-arm ledger	present in `zod-mirror-parity.test.ts`
the `::` row-key separator	56 occurrences — the spelling chosen to clear a neighbour's anchor
⭐ `handler-keys-string-any-mirrors-7344.test.ts`	still present, untouched

That last row is the one worth keeping. The first row-key spelling joined pair and key with a dot, which read as a property access on a mirror's handler-array key inside a file that already names that mirror — tripping objectui#7344's reader anchor, which pins that key's reader set at one file. It failed, correctly. The fix moved this PR's own separator rather than widening the neighbour's allow-list, so #7344 keeps its edge and this diff stayed at the two files the claim named. That is the declare-don't-absorb discipline paying out, and it was declared in the report before landing rather than discovered here.

What the card asked for, and what it turned out to be

All three acceptance boxes are delivered, and two of the three answers are not what the card expected:

The bucket is 25 keys, not 27. Re-derived two independent ways that agree — the ledger's own docblock classes, and a runtime walk marking every key whose slot reaches a recursion-breaking lazy node. My own review did not re-run those, but the report's method is stated and self-checking.
Zero re-filings — and that is the finding. The per-arm table covers all 52 ledger keys / 61 arms. Of the 9 union keys, exactly one sits in the SCHEMA-NODE bucket: objectui#7952's own, already ruled under objectstack#8593 and already repaired by PR fix(types): declare the component-node arm of DashboardComponentSchema.widgets on the TypeScript face #8296 (c842594). The other eight were each already CONCRETE in objectui#7759's open lane, arm by arm. ⇒ The finding(types): DashboardComponentSchema.widgets is a two-arm union in zod and a one-arm array in TypeScript — the ruled-legal metric-card component node parses green and tsc refuses it, and the parity ledger files the key as SCHEMA-NODE #7952 shape does not recur.
The granularity change and its pins landed as asked, with finding(types): DashboardComponentSchema.widgets is a two-arm union in zod and a one-arm array in TypeScript — the ruled-legal metric-card component node parses green and tsc refuses it, and the parity ledger files the key as SCHEMA-NODE #7952's key as the named positive control.

⭐ The systemic half — why a wrong number survived every green pin

WiderThanDeclared was the only one of the five ledgers in this file with no figure in its header. objectui#7733 pinned three ledgers' entry counts and objectui#8222 their key totals; neither reaches this one. So objectui#7759 published "27 keys / 21 pairs", this card's body inherited it, and the count went two out while every pin in the file stayed green.

A count with no pin is not a fact. It is a rumour with a citation.

The PR closes it: six figures for this ledger now sit in the header, read off WIDER_ARMS and the mirrors, and pinned — a hand-edited figure was one of the four ablation legs and it reddens.

⚠️ Third instance this shift of a ruling or card quoting a count that does not reproduce: 45→41 .default() sites on objectui#7735, 27→25 here, and 21→58 keys on objectui#7105's cast census. That pattern, not any one of the three, is what goes to the skills lane.

One correction I owe

My dispatch brief said "read PR #8296's two review comments". That PR has zero reviews — the two Clause-② reviews live on card objectui#7952 as comments 5568304348 and 5569569634. The dev read the right ones and recorded the correction rather than following a bad pointer. My error.

Bound recorded rather than left implicit

The census does not answer whether a single-arm SCHEMA-NODE key hides a second, concrete cause underneath the artifact — a different mechanism from hiding-behind-a-sibling, needing a two-face comparison per key rather than an arm enumeration. It is written into the ledger docblock, ⛔ not filed as a card, and that is the right call: it is neither a reproducible defect nor a contract breach.

Serial chain now released: objectui#7759 and objectui#7760 were held behind this card and are free to be dispatched.

Generated with Claude Code

https://claude.ai/code/session_01QtGhnU3WnnWyiWeYQhw2aX

Generated by Claude Code

fix(types): settle FilterField.operators, Container.maxWidth and Head…

Navigation Menu

Repository navigation

types: re-judge the 27 SCHEMA-NODE-bucketed keys of `zod-mirror-parity` per UNION ARM — a per-key verdict over per-arm facts let `DashboardComponentSchema.widgets`' concrete gap hide behind a schema-node sibling (census ordered by the #7952 ruling) #8252

Why

Scope

Acceptance

The coupling with #7105, declared rather than discovered

Adjudication — this card IS the ruling's second half

PM mechanism assumptions — measure these, and you are encouraged to falsify them

Gates

Route

ACCEPT — PR #8323 at head `afcd389`, `domain:spec @ objectui` seat, R2, 2026-09-07T12:40Z

Checklist conclusions

⭐ The finding inside the delivery — a stale figure that every green pin was blind to

⭐ Zero re-filings is the finding, not a skipped acceptance box

⭐ The declared collision — handled the way the round asked for

⚠️ My dispatch brief contained an error, and the dev caught it

Ablation and other evidence

Out-of-scope findings, verified as noted-not-filed

MERGED — PR #8323 landed as `645ecb9` on `main`, 2026-09-07T12:57Z

What the card asked for, and what it turned out to be

⭐ The systemic half — why a wrong number survived every green pin

One correction I owe

Bound recorded rather than left implicit

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

types: re-judge the 27 SCHEMA-NODE-bucketed keys of zod-mirror-parity per UNION ARM — a per-key verdict over per-arm facts let DashboardComponentSchema.widgets' concrete gap hide behind a schema-node sibling (census ordered by the #7952 ruling) #8252

Description

Why

Scope

Acceptance

Activity

os-zhuang commented on Sep 7, 2026

The coupling with #7105, declared rather than discovered

Adjudication — this card IS the ruling's second half

PM mechanism assumptions — measure these, and you are encouraged to falsify them

Gates

Route

claude commented on Sep 7, 2026

os-zhuang commented on Sep 7, 2026

ACCEPT — PR #8323 at head afcd389, domain:spec @ objectui seat, R2, 2026-09-07T12:40Z

Checklist conclusions

⭐ The finding inside the delivery — a stale figure that every green pin was blind to

⭐ Zero re-filings is the finding, not a skipped acceptance box

⭐ The declared collision — handled the way the round asked for

⚠️ My dispatch brief contained an error, and the dev caught it

Ablation and other evidence

Out-of-scope findings, verified as noted-not-filed

os-zhuang commented on Sep 7, 2026

MERGED — PR #8323 landed as 645ecb9 on main, 2026-09-07T12:57Z

What the card asked for, and what it turned out to be

⭐ The systemic half — why a wrong number survived every green pin

One correction I owe

Bound recorded rather than left implicit

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

types: re-judge the 27 SCHEMA-NODE-bucketed keys of `zod-mirror-parity` per UNION ARM — a per-key verdict over per-arm facts let `DashboardComponentSchema.widgets`' concrete gap hide behind a schema-node sibling (census ordered by the #7952 ruling) #8252

ACCEPT — PR #8323 at head `afcd389`, `domain:spec @ objectui` seat, R2, 2026-09-07T12:40Z

MERGED — PR #8323 landed as `645ecb9` on `main`, 2026-09-07T12:57Z