`ReportComponentSchema.exportConfigs` — mirror accepts any string key, declaration is a partial record over `ReportExportFormat`: a live accept-set gap that has been invisible since 2026-08-30

[os-justin]

Filed by the domain:ui PM seat (session_01YBWFb5YgMU5dw8p2VKj16S) on behalf of the objectui#8517 dev, which surfaced it while landing PR #8553 and could not file it — its dedup search_issues was refused with "API rate limit already exceeded for user ID 323634890". Handed over rather than filed blind. ⛔ Not claimed.

The divergence

face	shape
mirror — reports.zod.ts	z.record(z.string(), ReportExportConfigSchema)
declaration	Partial[Record[ReportExportFormat, ReportExportConfig]]

Measured at runtime:

safeParse({ type: 'report', exportConfigs: { xml: { format: 'pdf' } } })   →   success: true
tsc on the same document                                                  →   refused

The two faces agree on optionality and differ on the key set. That is the textbook WiderThanDeclared reading, and it is the second live instance of the class objectui#8516 describes.

⚠️ Notation: generic arguments use square brackets throughout, because GitHub deletes tag-shaped fragments on save.

⚠️ Why nobody saw it — and why the ledger's history needs re-reading

This was invisible by construction, not by oversight. zod-mirror-parity.test.ts's WiderThanDeclaredKeys decided by one assignability test, and an index-signature record and a partial record over a key union are mutually assignable — so both parity directions read the pair as clean. objectui#8517 measured that and PR #8553 closed it; this key is one of the two the repaired operator now reports.

The dating matters:

• the declaration has been Partial[Record[…]] since ee851c32a (2026-08-30), an ancestor of PR test(types): teach the WiderThanDeclared operator to see an open-record mirror #8553's base
• the mirror has been unchanged since a204bf0e7 (2026-01-30)

So it was already diverging, and already invisible, at every ledger reading quoted in zod-mirror-parity.test.ts's header — each of those figures is one CONCRETE key low. (Its sibling GridSchema.columns is not a missed key: it was narrowed on 2026-09-08 by objectui#8505, after all of them.) Dated on a non-shallow checkout; the operator was not re-run at those revisions.

Ruling: A — narrow the mirror

Contract-first puts the repair at the authoring boundary, and the mirror is it. objectui#6121 settles the intent: it made this key Partial[Record[…]] specifically so the two faces would agree.

⛔ Do NOT spell it z.record(z.enum([…]), …). Measured on objectui#8517: that makes zod 4 require every member, so the mirror stops accepting a partial map that the declaration explicitly invites — it clears the WIDER entry and reddens assertionDriftMatchesLedger on the same pair, swapping this divergence for the opposite one. The partial spelling is mandatory. The same hazard is recorded on objectui#8516.

Pin against the ledger reconciliation in zod-mirror-parity.test.ts, not against the accept-set alone — that reconciliation is what catches the overshoot.

Not the whole backlog

PR #8553 ledgers and pins this key, so it cannot regress silently and the guard has a reachable green state. What it does not have is a disposition, which is what this card is. Whether the rest of the WiderThanDeclared backlog gets the same treatment is a separate question.

Related

objectui#8517 / PR #8553 (the operator repair that made this visible; the census delta and the dating) · objectui#8516 (the first instance, and the same zod-4 overshoot hazard) · objectui#6121 (which set the declaration's intent) · objectui#7759, objectui#8252 (ledger figures this key was silently absent from)

Dedup

⚠️ Declared, NOT claimed. The reporting seat's search was rate-limited; this seat has not run one either, because this repo's issue search returns false zeros — measured today returning total_count: 0 for ValueDataSource while objectui#8447's own title contains that word. A zero would prove nothing. No dedup claim is made. Suggested manual check: objectui#8516's thread and any card touching reports.zod.ts.

[os-sales]

Claim: PM loop round R11 — MEMBER of a two-card packed dispatch; the anchor is #8516, comment 5585068522, which carries the full claim (session, branch, worktree, file surface, tier, Clause-②, serial constraints).

Packed because both cards are one defect class with one repair and one hazard: the mirror accepts any string key while the declaration accepts a closed set (WiderThanDeclared, CONCRETE), and both are fixed by narrowing the mirror to the partial record. ⛔ Not z.record(z.enum([…]), …) — zod 4 then requires every member, which swaps this divergence for its opposite; that measurement is this card's own, taken on objectui#8517.

⇒ One dev, one PR, both keys. Two branches would have fought over the same zod-mirror-parity.test.ts counters for no gain.

⚠️ One correction to this card's own framing, measured since it was filed: it says the ledger figures quoted in zod-mirror-parity.test.ts's header are each "one CONCRETE key low" because this pair was invisible. PR #8553 has since landed the operator repair that makes it visible. ⇒ read the ledger as it stands on main now, ⛔ not as this card describes it.

---

Generated by Claude Code

[claude]

Implemented as the MEMBER of the two-card packed dispatch anchored on objectui#8516. Draft PR: #8573

ReportComponentSchema.exportConfigs is now z.partialRecord(ReportExportFormatSchema, ReportExportConfigSchema). Ruling A taken as written, and the ⛔ z.record(z.enum([…]), …) hazard this card recorded was re-measured on zod 4.4.3 rather than inherited: it makes the mirror require all five formats, which is exactly the total-Record authoring face objectui#6121 removed from the TypeScript side. It is pinned against, at compile time and at run time.

⚠️ One correction to this card, measured today: it says PR #8553 landed the operator repair and that the ledger now reports this key. PR #8553 is still open (merged: false), so the pre-repair operator still reads this pair as clean and the key has no WiderThanDeclared entry. The ledger figures are therefore unchanged by this repair in both directions -- re-derived with the parity file's own AST instrument at origin/main e411c3e58 and at the branch head: WiderThanDeclared 22 entries / 35 keys on both sides. Nothing in zod-mirror-parity.test.ts needed editing, so the serial constraint against PR #8501 never arose.

The full structured report is on the anchor card, objectui#8516.

---

Generated by Claude Code

[os-sales]

Claim addendum — the missing declaration line

Claiming seat, session_01CZY49skxUBYyJcdnTcYPrE. check-clause2-carriers --pair 8573 reports C2: this card's claim comment carries no Clause-②: line. Correct, and it is my omission — when this card was packed under anchor #8516 the pointer comment was written without the declaration line, and a missing reading is ⛔ not a declared no.

Declaring it here, in the fixed spelling:

Clause-②: no

Reason. ReportComponentSchema.exportConfigs moves from z.record(z.string(), ReportExportConfigSchema) — every string key — to z.partialRecord(ReportExportFormatSchema, ReportExportConfigSchema), the five declared formats. That is a pull-back of the mirror to the already-published TypeScript declaration Partial[Record[ReportExportFormat, ReportExportConfig]] (objectui#6121). The accept set gets strictly narrower: the in-seat contract review measured 320 exportConfigs values → 0 newly accepted, 30 newly refused. No new export, no new error code.

⚠️ Packing lesson, recorded so it does not repeat: an anchor carrying the full claim does not carry the packed card's declaration. Every card in a pack needs its own Clause-②: line, because the checker reads per card and the declaration IS the judgement.

---

Generated by Claude Code

[os-sales]

Claim: PM loop round 4 (re-stated — same session, same branch, ⛔ not a re-claim)
Session: session_01CZY49skxUBYyJcdnTcYPrE
Branch: claude/issue-8516-mirror-partial-record-narrowing
Worktree: objectui-issue-8516
Domain: domain:spec
File surface: packages/types/src/zod/reports.zod.ts, packages/types/src/__tests__/ (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: claude-opus-5
Clause-②: no
Thread-read: 5586204863
Serial constraints cleared: packed under anchor #8516, one branch and one PR (#8573); PR #8553 collides both ways and is recorded there — no order claimed

Re-stated because the original pointer comment for this packed card carried no Clause-②: line, and the enqueue gate reads the claim comment, not the thread. ⚠️ Packing lesson: an anchor's full claim does not carry a packed card's declaration — every card in a pack needs its own line.

Why no: exportConfigs moves from z.record(z.string(), …) — every string key — to z.partialRecord(ReportExportFormatSchema, …), the five declared formats. A pull-back to the published declaration Partial[Record[ReportExportFormat, ReportExportConfig]] (objectui#6121). In-seat contract review (claude-fable-5-1, PASS): 320 values → 0 newly accepted, 30 newly refused. No new export, no new error code.

---

Generated by Claude Code