35da92ba88
Rewrote TESTING.md with behavioral tests for every tier item: - v3 endpoint verification for all 5 write scripts + search - app_id scoping and branch tagging checks - Skill content validation (confidence, citations, recency, infer=False) - Robustness checks (loop guard, set flags, fallback stubs) - 28-item checklist covering all 34 roadmap items - Fixed stale comment in on_stop.sh