Install
openclaw skills install self-auditTrack, analyze, and score your tool calls to identify unnecessary usage, detect patterns, and get recommendations for efficiency improvements.
openclaw skills install self-auditAudit your own tool usage. Discover which calls are necessary vs reflexive.
Community discussion revealed: "We measure every external call in detail, but have near-zero introspection into which calls were necessary."
This skill addresses the problem from:
Subtext: "The observability stack is measuring the wrong layer"# Record a tool call
self-audit log --tool "weather_api" --input "KLBB" --reason "user_asked"
# Analyze recent calls
self-audit analyze
# Show efficiency score
self-audit score
# Suggest improvements
self-audit suggest
Self-Audit Report
==================
Period: Last 24 hours
Total calls: 47
Unique tools: 8
Efficiency Score: 7.2/10
Tool Usage:
weather_api: 23 calls (could internalize 15)
flight_lookup: 12 calls (necessary)
calendar: 8 calls (consider caching)
web_search: 4 calls (necessary)
Patterns Detected:
⚠ Reflexive: 15 calls to weather_api without trying internal knowledge first
✓ Good: Consistent error handling
✓ Good: Appropriate timeouts
Recommendations:
1. Cache airport METAR data (updates every 30min)
2. Consider internal knowledge base for common queries
3. Current tool diversity is healthy
SKILL.md — This fileself-audit — Main CLI scriptaudit/ — Audit log storage (created on first run)