Trigger: any time after PR #8 merges (the findings it hands across live in that PR). Independent of the build.
Owner: Ian / the REMIT design team — findings flow to REMIT only as candidate register entries, never as silent coupling (ASSAY-DEC-3).
The four candidates (full entries with evidence and citations in docs/assay-findings.md §3):
- FIND-1 → audit REMIT shapes for stored midpoints/derived scalars; consider adopting ASSAY's representation-level ban (DEC-15) on top of NF10's behavioural one.
- FIND-2 (divergent — the strongest finding so far) → re-examine REMIT DEC-22's priority-weighted soft-commitment penalties at the argument surface: ASSAY re-derived the same least-worst mechanism with no weights at all (ordinal tiers + stated tie-breaks), so the weights are not load-bearing for the mechanism.
- FIND-3 → does REMIT's provenance model distinguish reported/assessed/assumption where planning machinery consumes them (ASSAY needed the four-way split, DEC-14)?
- FIND-4 (mild) → make REMIT's wall-clock exclusion structural (no wall-clock slot on stamped objects, by schema) rather than by kernel convention.
Also worth carrying across: the shared K2/DEC-16 dependency — both projects owe the same confidence→band-width derivation (findings ledger §4.4).
Trigger: any time after PR #8 merges (the findings it hands across live in that PR). Independent of the build.
Owner: Ian / the REMIT design team — findings flow to REMIT only as candidate register entries, never as silent coupling (ASSAY-DEC-3).
The four candidates (full entries with evidence and citations in
docs/assay-findings.md§3):Also worth carrying across: the shared K2/DEC-16 dependency — both projects owe the same confidence→band-width derivation (findings ledger §4.4).