OpenClaw agent health check and self-audit tool — 8-point system diagnostic.
OpenClaw Skill — works with OpenClaw AI agents
Performs an 8-point health audit of your OpenClaw agent: checks installed skill count, MEMORY.md length, cron distillation setup, daily log structure, core config file sizes, TOOLS.md configuration, workspace cleanliness, and facts.yaml health. Outputs a structured health report with actionable recommendations. Supports both full interactive audits and lightweight heartbeat-triggered quick scans.
# Install via ClawHub (recommended)
openclaw skill install agent-self-audit
# Or clone this repo into your skills directory
git clone https://github.com/rrrrrredy/agent-self-audit.git ~/.openclaw/skills/agent-self-audit- 8-point health diagnostic: Skills count, MEMORY.md length, cron distillation, daily log structure, SOUL.md/AGENTS.md size, TOOLS.md config, workspace cleanliness, facts.yaml health
- Traffic-light scoring: ✅ Healthy / 🟡 Needs attention / 🔴 Requires action for each check
- Smart skill cleanup: Identifies unused pre-installed skills (macOS-only, smart home, EU platforms, duplicate functionality)
- Memory optimization: Detects oversized MEMORY.md and suggests migration of details to facts.yaml
- Cron distillation audit: Verifies daily/weekly memory distillation jobs exist with proper constraints
- Heartbeat lightweight mode: 3-item quick scan (MEMORY.md lines, skill count, facts.yaml existence) for periodic automated checks
- Safe-by-default: All write operations require explicit user confirmation before execution
"自检" → Full 8-point health audit
"系统优化" → Full audit with optimization suggestions
"健康检查" → Full audit
"检查龙虾状态" → Full audit
"优化记忆" → Memory-focused audit
Add to HEARTBEAT.md for weekly automated checks:
- agent-self-audit 周检:当前为周一且本周尚未执行自检时,自动运行全量健康检查,输出报告
Or via cron:
openclaw cron add --name "agent-weekly-audit" --cron "0 1 * * 1" \
--message "执行 agent-self-audit 健康检查"agent-self-audit/
├── SKILL.md # Main skill definition
├── scripts/
│ ├── health_check.py # Cross-platform quick 3-item health check script
│ └── health_check.sh # Bash quick 3-item health check script
├── references/
│ └── audit-criteria.md # Detailed check criteria
└── .gitignore
- OpenClaw agent runtime
- Python 3 (for inline diagnostic scripts)
- Standard Unix tools (
wc,ls,find,grep)