Deep FSV Verification

Use this manual skill only when explicitly requested, when a specialist escalates uncertain source-of-truth evidence, or when a high-risk mutation crosses security, risk-control, SQLite, prompt-safety, or release-gate boundaries.

Load references/adversarial-checklist.md for the full heavyweight checklist. Routine implementation should use the slim fsv-verify doctrine skill first.

Minimum output still follows .claude/rules/evidence-schema.yml and must include PRE source-of-truth read, ACT, POST source-of-truth reread, expected DIFF, and HALT/GO decision.