agent-health-optimizer_skill

Overview

This skill automates diagnostics and continuous self-optimization for an OpenClaw agent. Run a single command to get a 0–100 health score with a letter grade, detailed audits for memory and cron jobs, and a skills comparison against the ClawHub catalog. It can be scheduled weekly to track trends and surface prioritized repair items.

How this skill works

The toolkit runs four analyzers: health scoring, memory auditing, cron inspection, and skill comparison. It reads workspace files, logs, skill metadata, and the OpenClaw cron list, then writes structured JSON reports into memory/ for trend tracking. An optional --fix mode safely backs up cron config before applying automated cron repairs.

When to use it

• Diagnosing unexplained agent failures or degraded behavior
• Preparing an agent for production or handoff
• Checking installed skills against ClawHub for upgrades or missing categories
• Automating weekly health checks and trend tracking
• Before applying broad configuration changes or infrastructure updates

Best practices

• Run the full self_optimize runner weekly in read-only mode; review suggested fixes before applying --fix
• Keep python3 and the openclaw CLI up to date for accurate cron analysis
• Back up your workspace and review memory/*.md files regularly to reduce stale entries
• Use isolated session cron runs for read-only audits to avoid side effects
• Apply --fix only after reviewing the cron-backup.json created by the tool

Example use cases

• One-command agent triage: run self_optimize.py to get a prioritized repair list and health grade
• Memory hygiene sweep: detect injection risks, missing source tags, and stale facts with memory_auditor.py
• Cron remediation: find schedule collisions, add stagger/jitter, and disable duplicate announcements with cron_optimizer.py --fix
• Skill modernization: compare installed skills to ClawHub to find higher-rated alternatives and gaps in category coverage
• Automated monitoring: add a weekly read-only cron to produce trend reports and detect regressions early