finding(plugin-form): three of ObjectForm's four section maps read `visibleWhen` through an unnecessary `as any` — found twice by two rounds, filed by neither

[os-sam]

Filed by the domain:ui execution seat (PM session session_013hfmP9hoMd3dJwTh85J4yB). ⛔ Recording only — not graded, no domain:*, finding only; grading is triage's.

Filed because two independent dev rounds found this and neither filed it, each for the same structurally self-defeating reason: it is three lines in a file that round did not hold. That reasoning is individually correct and collectively guarantees the finding is never recorded — which is the orphaning pattern this lane has been tracking all day.

The reading

packages/plugin-form/src/ObjectForm.tsx rebuilds the section list once per layout route. Four maps; the predicate is read through an unnecessary as any in three of them:

map	site (re-derive on your own base)	cast
split	~:350	as any
drawer	~:383	as any
modal	~:414	as any
tabbed	~:292	none

⚠️ Line numbers are approximate and taken from two reports on different bases — re-derive them. The file has changed under both PR #6791 and PR #6903.

ObjectFormSection.visibleWhen is declared in packages/types, so the cast buys nothing. What it costs is the compiler: with as any in the path, a rename of that key would silently copy undefined at those three sites instead of failing the build.

⇒ That is precisely the silent-drop failure mode objectui#6237 exists to close, sitting latent in the file that card's own fix passes through.

Why this is worth a card despite being three lines

⭐ The correct form is already in the same file. The tabbed map at ~:292 was written without the cast, deliberately, with a comment recording why — PR #6791's dev removed it after a per-rule lint baseline caught the cast adding a 57th no-explicit-any to the file. So this is not a design question; it is three lines converging on a pattern their own sibling already demonstrates, four lines away.

⛔ No behaviour change and no measured defect today. The key is copied correctly at all four sites right now. The exposure is entirely prospective: it is the compiler's ability to catch a future rename.

Provenance — two independent sightings

1. PR feat(plugin-form): honour a section visibleWhen on the tabbed layout arm #6791 (round 3 of The tabbed modal arm (ModalForm contentLayout: 'tabbed') cannot carry a section visibleWhen at all — FormFieldTab declares no predicate slot #6237) recorded it in its report's out_of_scope_findings and declined to file: "three lines in a file I hold this round for a different reason, with no behaviour change and no measured defect; folding it in would be unreviewable scope creep, and filing a card for a cast is dispatch overhead."
2. PR fix(plugin-form): declare WizardStepConfig independently of FormSectionConfig #6903 (round 4 of The tabbed modal arm (ModalForm contentLayout: 'tabbed') cannot carry a section visibleWhen at all — FormFieldTab declares no predicate slot #6237) found it again independently and declined again, adding a second reason: "it is three lines in a file I do not hold this round at all (my diff does not touch ObjectForm.tsx), and the bounded-in-place exemption fails on the 'declared file face' condition."

⇒ Both refusals are correct on their own terms. ⛔ The point of this card is that the conjunction of two correct refusals is a permanent gap: a finding whose only plausible filers are seats structurally disqualified from filing it.

Suggested disposition (not a ruling)

Cheapest sound close: fold the three casts into whichever card next legitimately holds ObjectForm.tsx, rather than dispatching a card for three lines. ⚠️ If triage grades it standalone, note that the whole value is compile-time — there is no runtime assertion that can pin it, so a reviewer should expect a type-level check or nothing.

Related: objectui#6237 (the family and the fix that runs through this file) · PR #6791 · PR #6903