hook-audit
Installation
SKILL.md
Audit ~/.claude/hook-metrics/ using REFERENCE.md for metrics, cohorts, and data gaps.
Modes: default/--hooks activity; --retro adds flow and environment findings; --all includes retro, latency, skill firing, drift.
Flow
- Inventory hooks/date range; separate evals from real runs. Cohort by harness version/model; split or exclude mixed-model sessions using
model-switches.jsonl. - Aggregate blocks, warnings, nudges, denials, sessions, trend; compute P50/P95 and wall time when requested.
- Compare scripts, observed keys, and enforcement; distinguish true zero-fire candidates, untested hooks, and advisory rules.
- Retro: follow Session-environment retrospective; add available retro metrics. Missing telemetry does not block transcript findings.
- All: inspect
skill-fires.jsonl; runbash scripts/generate-hook-configs.sh --check. - Model-switch policy: use
/quantify-impact; success/rework is primary, cache-write cost a guardrail. - Rank at most five actions across telemetry and session findings by impact. Recommend only unless implementation is requested.
Before deletion, shadow via HOOK_SHADOW_RULES in a representative, version-qualified trial; compare outcomes/violations. Never shadow strict safety/permission.
Done
Report available metrics/values, sample size, 7-day trend, next action; mark gaps unavailable. Below five comparable real sessions is preliminary. Cite source files and exact harness_version + model cohort for prune/severity changes.