Motir

Found while running MOTIR-3947 (motir-core#2472). The card's core claims held — the 69-rule count is exact and the defect reproduces — so this is telemetry about the AUTHORING pass, not a blocker on anything.

MOTIR-3947 was carved out of MOTIR-3933 by the re-plan that card's own run submitted. Two claims rode across the split without being re-verified against shipped code.

1 — the consumer named in acceptance criterion 2 does not exist in that form

"The ADR is checked against BOTH known consumers — the onboarding Style gallery and TokensSpecimen's galleries"

At origin/main (0068da2f4) there is no onboarding Style gallery. components/onboarding/DesignStep.tsx renders a chip-row StylePicker (no vignettes) and one LIVE-mode StyleVignette at line 196, inside a <section> that itself carries data-style + data-appearance-scope. It is a single-tile nested scope, and it IS affected — a Motir user whose own appearance is a material style sees the wizard's preview wearing Motir's material rather than the project's — but a runner following the criterion literally looks for a gallery, does not find one, and has to decide between inventing it and reporting a false gap.

The two actual scoped-style galleries are packages/design-system/src/specimen/TokensSpecimen.tsx:103 and app/tokens/page.tsx:349. The card names neither /tokens nor the onboarding page's real shape.

The phrase is traceable: MOTIR-3947's own body says "motir.co/design, and the onboarding Style gallery has the same shape", and StyleVignette's header describes SCOPED mode as "the onboarding Style gallery (one vignette per style)" — a comment written when 7.3.37 landed and never re-checked against what DesignStep actually renders today. The card inherited the comment, not the code.

2 — the aurora figure is one tile optimistic

The card's table reads aurora → 3 / 11. Re-measured on origin/main with a 12-property × 7-element probe: 2 / 11. The extra failing tile is neumorphism, differing in exactly three background-image declarations and nothing else, so a narrower probe scores it a pass.

This is probe breadth, not drift — same defect, same ref, stricter instrument — and it is recorded because the direction is worse, which is the direction that arrives with an innocent explanation already attached. The card also measured 3 of 11 possible ancestors; the full matrix is 64 / 121, with six styles leaking rather than the two the table shows.

What actually went wrong

Neither claim was cheap to check and both were checkable in one command. The lesson that describes it is the global "A split or renumber does not inherit prose corrections" — reinforced against this card (recurrenceCount 2). The authoring-side gate is precondition-verification: a consumer a card instructs a future runner to check is a precondition, and a count carried across a split is a measurement whose ref moved.

Disposition — already applied, nothing outstanding

• The ADR (docs/decisions/scoped-preview-isolation.md, §Q6) covers all three real consumers plus the three LIVE-mode call sites, and carries the correction as a marked note rather than silently satisfying the criterion.
• §Q1 carries both re-measurements with their commands, and the full 11 × 11 matrix.
• Both are stated in motir-core#2472's body.
• MOTIR-3947's own comment thread records the correction against the card.

No amendment was made to MOTIR-3947's description. Rewriting an acceptance criterion to match what the run did is the shape this corpus calls a silent re-scope; the card stands as authored and the correction sits beside it.

Sibling record

MOTIR-3935 is the planning bug about the same authoring pass's SIZING. This is the same pass's prose, and the two are separate findings of one origin.