weekly-self-improve-loop

What This Skill Does

The weekly-self-improve-loop is an automated diagnostic and growth tool for OpenClaw agents. It functions by periodically auditing the agent's memory logs from the previous seven days, calculating performance metrics such as task completion rates and blocker frequency, and identifying recurring friction patterns. By analyzing these data points, the skill proactively generates new recovery skills or updates existing ones, ensuring that the agent's autonomy baseline improves over time rather than stagnating. This creates a self-healing loop where the agent learns from its own failures.

Installation

To install this skill, use the ClawHub command within your terminal or agent interface: clawhub install openclaw/skills/skills/dalomeve/weekly-self-improve-loop

Use Cases

• Continuous Improvement: Perfect for long-term agents managing complex, multi-week projects where minor operational inefficiencies might otherwise accumulate.
• Bottleneck Identification: Ideal for developers or project managers who need to know exactly why tasks are stalling without manual log review.
• Automated Skill Building: Useful for maintaining a library of specialized 'recovery' skills that address common errors like network timeouts or API rate limits automatically.

Example Prompts

1. "OpenClaw, perform the weekly-self-improve-loop now to see where I've been hitting blockers this week."
2. "Run the self-improvement audit and show me the top 3 friction patterns you found in the last week."
3. "Could you trigger a manual review of my memory files and generate new recovery skills based on recent task failures?"

Tips & Limitations

• Frequency: While designed for a Sunday cadence, you can trigger this manually after any high-intensity project to capture lessons learned while they are fresh.
• Data Privacy: This skill is strictly local-first. It processes memory logs directly on the host machine, ensuring your sensitive project data is never shared externally.
• Pattern Quality: The efficiency of this skill depends on the quality of your task logging. Ensure that blockers are clearly marked with the 'Blocker:' prefix for the regex to capture them correctly. If you do not tag your memory files consistently, the report may lack actionable data.