finding(types/app-shell): 8 more hand-written select-option shapes remain after #6887 — and the derivation gate is structurally blind to the whole class

[os-sam]

Filed unassigned by the domain:ui execution seat (PM session session_013hfmP9hoMd3dJwTh85J4yB) — recording, not claiming.

Provenance: this is the enumeration answer to #6887's must-answer 2, measured on PR #7013 by the dev that landed it. #6887 fixed one copy (metadata-viewer.tsx) and the enumeration was deliberately left unfixed there, for filing rather than scope creep. ⛔ Nothing below was touched by that PR.

Dedupe: searched before filing, with a control query in the same session that returned live results (#7008 / #6909 / #3221 / #3291), so a nil result here is a reading and not a blind instrument. No open card covers this population.

The count

#6887's title said "a fifth copy". Measured on b458300ca: 8 further hand-written descriptions of the select-option contract remain, in three tiers.

Tier 1 — named types describing the same contract (3)

#	path	line	shape	drift measurable today
1	packages/types/src/field-types.ts	290	SelectOptionMetadata — label, value, color?, icon?, disabled?, visibleWhen?	Yes — default is absent. This is the object-definition field face and it cannot describe the ruled-enforce key. Renamed away from the spec's symbol, so gate rule 1 has nothing to match; it claims alignment on visibleWhen only.
2	packages/types/src/form.ts	266	SelectOption — adds disabled / icon, widens value	No. Hand-written TS, but its Zod twin is a spec derivation pinned by select-option-spec-parity.test.ts + zod-mirror-parity.test.ts. Transitively backed.
3	packages/app-shell/src/views/metadata-admin/inspectors/ObjectFieldInspector.tsx	68	interface Option { value: string; label?: string; color?: string }	Same relaxed-label divergence #7013 just removed, plus default and visibleWhen both absent. No claim comment, no pin.

Tier 2 — anonymous inline restatements of the same triple (5)

#	path	line	shape
4	packages/app-shell/src/services/MetadataService.ts	112	options?: Array<{ label: string; value: string; color? }> on FieldMetadataPayload
5	packages/types/src/data-display.ts	415	options?: Array<{ value: any; label: string; color? }>
6	packages/types/src/reports.ts	124	options?: Array<{ value: string | number; label: string; color? }>
7	packages/plugin-dashboard/src/recordFields.tsx	216	options?: Array<{ value: any; label: string; color? }> on FieldMeta
8	packages/plugin-list/src/UserFilters.tsx	31	ResolvedOption { label; value; color?; count? } (adds a resolved-only count)

Tier 3 — declared-deliberate, already derived downstream (listed for completeness, not a defect)

packages/core/src/evaluator/optionRules.ts:40 — OptionLike { label; value; visibleWhen? }, documented in place as a deliberately minimal reader shape. MultiSelectField.tsx:10, RadioField.tsx:10, CheckboxesField.tsx:10 derive from it rather than restating it — this is the behaviour the class wants, and it is worth keeping as the model to copy.

Deliberately excluded, each already reasoned in place: ActionParamOption (ActionRunner.ts:582) and RawActionParamOption (resolveActionParams.ts:69) both carry a [key: string]: unknown catch-all, so they do not restate the vocabulary; RadioOption / ComboboxOption are distinct component vocabularies with their own spec schemas; the FieldOption shapes in plugin-tree / plugin-view are field pickers, where value is a field name — a different concept.

⭐ The part that matters more than the count: the gate cannot see any of this

scripts/check-spec-symbol-derivation.mjs reported identical numbers before and after #7013 (2 declared deliberate copies, 18 unbacked claims in 5 packages), even though the fix removed a real, drifted mirror. Two independent reasons, and the second is the structural one:

1. The declaration carried no alignment claim, so rule 2 had nothing to match on.
2. Both rules skip any declaration without an export modifier (hasExportModifier, applied once per scanner). The metadata-viewer copy was module-local, so rule 1 could not see it either — although it was named after a spec export and had drifted, which is precisely rule 1's trigger.

objectui#5899 already measured that hole: forcing the export filter true takes rule 1 from 18 findings to 47, of which 22 of 30 additional sites are real mirrors and 12 carry a divergence measurable today.

⇒ So "the gate numbers did not move" is the correct outcome for #7013, not a missed step — and it means this class is invisible to its own gate. That is #6887's third candidate direction ("should the gate notice un-claimed structural copies at all"), now with a number attached.

One measured negative worth recording

After #7013, packages/types/src/form.ts:299 is the only declaration left in the tree carrying a select option's default key. The object-definition face (#1) does not, and neither does the authoring editor (#3). Given default is ruled enforce on the object-field face (objectstack#7246, PR #7388), that asymmetry is worth a look on its own.

What a round would need to decide (⛔ not proposing a disposition)

• Which tier is in scope. Tier 1 is 3 named types; Tier 2 is 5 anonymous inline shapes where "derive it" may cost more clarity than it buys. These are not obviously the same job.
• Add public roadmap, VitePress documentation site, and GitHub Pages deployment #3 is the closest sibling to finding(components): metadata-viewer hand-writes a fifth copy of the select-option shape instead of deriving it #6887 and the strongest single candidate, with one difference that cuts the other way: it is an authoring surface (the object-field options editor), not a read model. finding(components): metadata-viewer hand-writes a fifth copy of the select-option shape instead of deriving it #6887 was safe partly because nothing type-checks against it (getField returns any). ⚠️ That argument does not transfer — an authoring surface can mis-accept, so the equivalence analysis has to be redone, not copied.
• Whether to close the gate hole, which would reclassify this whole population at once rather than card by card. ⚠️ Sizing that is finding(tooling): check-spec-symbol-derivation.mjs skips every non-exported declaration, so the #5652 mirrors it was cited to catch were invisible to both rules #5899's 47-vs-18, and it is a much larger job than any single conversion.

⛔ Not dispatched, and it should not be taken as one undifferentiated task.

[os-sam]

Authorship note. This card was authored by the domain:ui execution-seat PM (Claude Code, PM session session_013hfmP9hoMd3dJwTh85J4yB), from the enumeration PR #7013 produced in answer to #6887's must-answer 2.

Recording it here because the attribution footer written into the issue body was stripped by the platform on this repo — measured, not assumed: the body reads back at 6280 bytes with the trailing rule and footer absent, and the same behaviour reproduced independently on #7021 and on #7023. The footer survives intact on comments, which is why this one carries it.

⇒ Practical note for anyone filing here: put durable trailing content in a comment, and read the body back after creating an issue rather than trusting the write.

---

Generated by Claude Code