Implementation arc spawned by the #773 root-cause analysis (docs/defect-shapes-773.md). Detector-first, ordered by recorded coverage: #774 (set-equality guards, 39% of failures) and #775 (guard mutation proofs) are the load-bearing pair; #776 makes hook execution observable so the rest can be measured at all. Then #777 (LSPs), #778 (check-and-use races), #779 (fail-open + destructive replace), #780 (Python tooling), #781 (config hygiene), #784 (no-session-narrative), #794 (witness a fix's test failing before the fix).
ENFORCEMENT FOLLOW-ONS, added 2026-08-21 after #778 showed the rules are stated but not machine-checked: #785 (mutation proofs for the unproven guards), #790 (an executable clause-level mutation harness, which also verifies that a MUTATION row's claim is TRUE — the residue #775 admits), #806 (guard populations over FILES derive from the git index, not a filesystem walk). These were outside the milestone while being exactly what converts 'review caught it' into 'the suite caught it'; #778 needed fifteen review rounds and reproduced four of this analysis's own defect shapes in its deliverable, which is the evidence for pulling them in.
Members are the real pickups; this milestone is the ranking mechanism, not a status list — live state is the issues themselves.
Claude Design ↔ shipped SPA sync + design-driven feature work. Repurposed from the completed "Design-sync workflow ready" arc.