agentsclimarketplace

Hook audit

Skill armueller/codebase-guardian/skills/hook-audit

Audit validation hook performance — analyze allow/deny decisions, timing, and false positivesFrom its SKILL.md

Install
npx -y skills add armueller/codebase-guardian --skill hook-audit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.7 KB, as published. Nobody here has run it

Validation Hook Performance Audit

You are analyzing the Codebase Guardian validation hook's performance for this project.

Step 1: Locate the Log File

PROJECT_ROOT=$(git rev-parse --show-toplevel 2>/dev/null || pwd)
PROJECT_HASH=$(echo -n "$PROJECT_ROOT" | shasum -a 256 | cut -c1-12)
# Guardian's data now lives in the plugin's data dir. Prefer an explicit override,
# then ${CLAUDE_PLUGIN_DATA} (substituted here when installed as a plugin, or read
# from the environment), then the deterministic plugin data path as a guaranteed fallback.
GUARDIAN_HOME="${GUARDIAN_HOME:-${CLAUDE_PLUGIN_DATA:-$HOME/.claude/plugins/data/codebase-guardian-codebase-guardian}}"
LOG_PATH="$GUARDIAN_HOME/logs/$PROJECT_HASH/validation-debug.log"

If the log file doesn't exist, inform the user:

No validation logs found for this project. The hook hasn't run yet, or logs have been cleared.

If there's also a .old rotated log, read both (current + rotated) for a fuller picture.

Step 2: Parse Log Entries

Each hook invocation starts with === {timestamp} === and contains structured entries:

Key patterns to extract:

  • ALLOW: {reason} — Edit was allowed
  • DENY: {reason} — Edit was blocked
  • ERROR: {message} — Hook encountered an error
  • [TIMING] TOTAL VALIDATION TIME: {N}ms — Total time per validation
  • [TIMING] Headless Claude execution: {N}ms — AI validation time
  • [TIMING] Build pattern context: {N}ms — Code index query time
  • [TIMING] Cache hit / [TIMING] Cache miss — Cache effectiveness
  • [SESSION] Retry attempt — Developer is retrying after denial
  • [SESSION] Identical resubmission — Developer resubmitted without changes
  • [SESSION] File content changed since last denial — Stale session cleared
  • Decision: allow / Decision: deny — Headless Claude's decision
  • Violations: — What violations were found

Step 3: Generate Report

## Codebase Guardian — Hook Performance Audit

**Project:** {project name}
**Log period:** {first timestamp} to {last timestamp}
**Total validations:** {N}

### Decision Summary

| Decision | Count | Percentage |
|----------|-------|------------|
| Allow | X | XX% |
| Deny | X | XX% |
| Error (fail-open) | X | XX% |
| Cache hit (skipped) | X | XX% |
| Timeout (fail-open) | X | XX% |

### Timing Distribution

| Metric | p50 | p90 | p99 | Max |
|--------|-----|-----|-----|-----|
| Total validation | Xms | Xms | Xms | Xms |
| Headless Claude | Xms | Xms | Xms | Xms |
| Pattern context | Xms | Xms | Xms | Xms |

### Session Analysis

- Retry attempts: {N} (developer revised and resubmitted)
- Identical resubmissions: {N} (developer resubmitted without changes — wasted time)
- Stale sessions cleared: {N} (file changed between retries)
- Average attempts before allow: {N}

### Denial Analysis

**Most common violation types:**
1. Missing JSDoc — {N} occurrences
2. DRY violation — {N} occurrences
3. Pattern mismatch — {N} occurrences
4. Inline comment quality — {N} occurrences

**Files most frequently denied:**
1. `path/to/file.ts` — {N} denials
2. ...

### Potential False Positives

Look for patterns that suggest the hook is being too aggressive:
- Functions denied multiple times for the same file in the same session (developer struggling to satisfy the hook)
- Denials followed immediately by identical resubmissions (developer may not understand the violation)
- Denials where the violation text is vague or unhelpful

List any suspicious patterns found.

### Improvement Recommendations

Based on the analysis, suggest:
- If too many timeouts: consider increasing timeout or reducing context size
- If too many false positives: identify which rules are triggering incorrectly
- If cache hit rate is low: check if edits are highly varied or cache TTL is too short
- If retry rate is high: check if violation messages are actionable enough

Step 4: Offer Actions

  1. Clear old logs — Offer to remove the rotated .old log if it's stale
  2. Adjust hook configuration — If patterns suggest config changes (timeout, enforcement mode)
  3. Report false positives — Help the user file specific false positive examples for future tuning

Important Notes

  • Parse the ENTIRE log file, not just a sample
  • Timing percentiles should be calculated from actual data, not estimated
  • Be specific about false positive candidates — include the actual violation text and the code that was denied
  • If the log is very large (>10MB), focus on the most recent 1000 validations and note the sampling

Keep looking

Skills are one crate of 608,475. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.