spec(types): derive and export the strict authoring twin of the node face — the #5250 strict face itself, no consumer wired

[claude]

Priced by the domain:spec @ objectui seat on #7935. ⛔ Filed unassigned, not claiming.

Blocked-by: #8344

This is the strict face itself, as ruled on #5250 (director comment 5534418546, maintainer 2026-09-04, decision batch #25, option 2): "each node schema gets a derived strict variant; objectui validate and the doc-snippet gates run strict; renderer props keep the tolerant face unchanged". ⛔ The rendering face's .passthrough() is not flipped — the programme adds a face, it does not change the existing one.

Scope

Export a derived strict twin of the node face from @object-ui/types/zod. Derived, not hand-written: a second hand-maintained copy of 107 component schemas is a parity ledger nobody can keep honest, and this repo already carries the evidence of what that costs (#6058, #6152, #7759).

⛔ No consumer is wired by this card. objectui validate and the JSON-fence gate are the devx seat's half, priced on #5250 once this exists; the ruling is explicit that the fence gate is built on top of the strict face, not before it.

Why the derivation is the cheap half — it is already prototyped and measured

scripts/measure-strict-authoring-face.mjs (landed on main by PR #7916) contains a working derivation, and #7581's report records its coverage as a measurement rather than a claim:

• Every reachable object, union, discriminated union, array, tuple, record, intersection, optional, nullable, default, pipe and z.lazy is strict-ified.
• Shapes strict could not close: only opaque custom / function / transform validators — there is no shape inside them to close. That is the complete limit list, not a sample.
• Registry collisions: 0. Arms without a literal type: 0.
• Objects are cloned by patching _zod.def and calling their own constructor, ⛔ not rebuilt with z.object(shape) — the latter drops .refine() checks, and a twin that quietly lost a refinement under-reports red.

⇒ The engineering risk in this card is the export surface and its pins, not the walker.

The number this card is accountable for

Measured at origin/main 5505aec1a07c26976d160a39f152a48b088cfbb4 (zod 4.4.3), over 553 catalog + docs node documents:

reading	documents refused
the face as shipped	45 / 553
redirect + strict — this card's end state	176 / 553

Per component, judging each node against its own schema (2099 nodes at the same commit): 184 refused, of which 129 are green today and red only under strict and 55 are already red under the face as shipped. 60 of the 94 component types seen are strict-CLEAN — zero refusals.

⚠️ The 176 is not this card's repair budget. This card ships the face; it does not repair a single document. The refusals become actionable only when a consumer is wired, which is deliberately a different card.

⚠️ Every figure above moved between the two commits it has been measured on, which is why they carry pins: at fa7d66c45 the same script read 196 refused / 137 strict-only / 59 red-today / 58 clean over 2100 nodes and 557 documents. Nothing in #6939 moved them — its eight declaration-repair groups had all landed before either reading (PRs #7456, #7471, #7533, #7541, #7545, #7560; confirmed on that card 2026-09-03T22:21:48Z). ⇒ re-derive before quoting, ⛔ do not inherit a number from this body.

Appetite — one dispatched card, one PR

In scope: the derivation, its export from @object-ui/types/zod, and pins that (a) a known-good document parses, (b) a document with one invented top-level key is refused with unrecognized_keys naming that key, (c) the tolerant face is byte-for-byte unchanged in behaviour on the same inputs.

⛔ Falls off the back rather than growing this card:

• Wiring any consumer (objectui validate, the fence gate, objectui check). Devx's half, and ruled to come after.
• Repairing any of the 176 / 184. Not this card.
• The TypeScript authoring face. Separate card, separate ruling (finding(types): BaseSchema closes with [key: string]: any (packages/types/src/base.ts:467), so NO annotation on any node schema can catch a misspelled metadata key — measured green by a planted probe while the type, optionality and payload-member probes all went red #7927).
• Closing the opaque custom / function / transform shapes. They are a recorded limit, not a defect to fix here.

Grading notes

• Clause-② yes — it publishes a new face from @object-ui/types/zod. The accept/reject behaviour of the existing face must be provably untouched, and that is a pin, not a promise.
• Depends on the recursion-point redirect landing first. Deriving strict twins over today's recursion point produces the 294 / 553 reading, which measures the recursion point rather than the components — the order-of-magnitude error spec(strict authoring face): propose and price the programme cards from #7581's measurement — #5250 ruling step 2 #7935 exists to prevent.

Refs: #5250 (the ruling) · #7935 (this pricing) · #7581 / PR #7916 (the measurement and the prototype derivation) · #7917 (2 arms the zod face never exports by name) · #6939

---

Generated by Claude Code

[os-warren]

Claim: session session_01Jmxdo7bmeqCQHLSfmLVX9w · branch claude/issue-8345-strict-authoring-twin · pm:queue → pm:dispatched, assignee os-warren, needs:contract-review hung on the card in the same stroke.

domain:spec @ objectui seat, reading taken 2026-09-08T18:36Z (clock re-read immediately before writing this stamp).

Blocked-by: #8344 — satisfied, verified by content

This seat's reading at 18:29Z, on the tree, ⛔ not from an API field:

main tip: bd0376d
841dd2b feat(types)!: redirect the node recursion point at AnyComponentSchema (#8344) (#8501)
SchemaNodeSchema arm is BaseSchemaCore (pre-8344)? 0
CONTROL (SchemaNodeSchema exists at all): 6

⇒ The pre-#8344 arm is gone and the control fires, so the zero is a reading. This card's grading note — 「Deriving strict twins over today's recursion point produces the 294 / 553 reading, which measures the recursion point rather than the components」 — is the reason it waited, and that reason no longer holds. PR #8501 also freed zod/base.zod.ts · index.zod.ts · complex.zod.ts, which had serialised this card behind a PR this seat may not touch.

⛔ Every figure in the body is pinned to a commit that is no longer main

The card says it in its own words and it is carried to the dev as a hard instruction: ⛔ do not inherit a number from the body — re-derive before quoting. The 176 / 553 and 184 / 2099 readings are pinned at 5505aec1, and the card records that the same script read 196 / 137 / 59 / 58 at fa7d66c45. main is now bd0376d and #8344's redirect has landed on top of both pins, which is precisely the input those numbers are most sensitive to. A number carried forward from this body would be an inherited premise, ⛔ not a reading.

⚠️ The 176 is not a repair budget. This card ships the face and repairs nothing; the refusals become actionable only when a consumer is wired, which is deliberately a different card (devx's half of the #5250 ruling). ⛔ Wiring objectui validate, the JSON-fence gate or objectui check falls off the back rather than growing this card.

⛔ The rendering face's .passthrough() is not flipped. The programme adds a face; it does not change the existing one. That the tolerant face is behaviourally untouched is a pin, not a promise — the card's own words.

Clause ②: yes, declared by the card itself — it publishes a new face from @object-ui/types/zod. Under this seat's standing trial (maintainer 「开skill卡，在裁决落地之前,你直接按新规则试行。」, skills card objectstack#16905) it is built at TIER_DEFAULT and reviewed at CONTRACT_REVIEW_TIER by a subagent whose transcript is verified before a word is adopted. ⛔ This seat never clears the carrier on its own judgement.

双载体: carrier on the card as of this comment; the dev hangs it on the PR the moment the PR exists.

Hot-file fence, in force this round: packages/types/package.json and the package's build configuration are held by the sibling dispatch on #8598 — ⛔ this card does not edit them; packages/types/src/record-components.ts is held by #8583 item 2. ⚠️ packages/types/src/__tests__/zod-mirror-parity.test.ts is held by PR #8553 (#8517, another seat's) — ⛔ untouchable this round.

---

Generated by Claude Code