self-improvement

What This Skill Does

The self-improvement skill for OpenClaw is a systematic framework designed to turn every interaction, error, and success into a persistent asset for your AI agent. By maintaining a structured directory of logs (.learnings/), the agent creates a feedback loop where it learns from its own history. This skill transforms the agent from a stateless assistant into an evolving entity that avoids repeating mistakes and adopts refined best practices over time.

Installation

To integrate this skill into your environment, you can use the official ClawdHub CLI:

clawdhub install self-improving-agent

Alternatively, for a manual setup, clone the repository directly into your local OpenClaw skills directory:

git clone https://github.com/peterskoett/self-improving-agent.git ~/.openclaw/skills/self-improving-agent

Ensure that you have initialized the necessary workspace structure by creating the .learnings directory inside your project root:

mkdir -p ~/.openclaw/workspace/.learnings

Use Cases

This skill is essential for complex development workflows and long-term project management. Use it when:

• An API call or CLI command fails, allowing the agent to document the error for later root-cause analysis.
• The user corrects the agent, ensuring the correction is stored in LEARNINGS.md so the agent improves its future accuracy.
• You identify a missing feature, logging it in FEATURE_REQUESTS.md to keep the roadmap visible.
• You discover a more efficient coding pattern, which is then codified as a 'best_practice' for the entire project.
• You realize your knowledge is outdated, flagging it as a 'knowledge_gap' to prompt external research or updates.

Example Prompts

1. "Actually, the API endpoint changed last week to v2, please remember this for future requests." (Agent logs as correction)
2. "That command failed with a permissions error, please log this in ERRORS.md and suggest a sudo-based workaround."
3. "We should adopt the new modular structure we just discussed as a project-wide standard. Add this to LEARNINGS.md as a best_practice."

Tips & Limitations

To maximize the utility of this skill, regularly perform a 'promotion' of learnings. Don't let your logs become a graveyard of information; move stable, recurring patterns from LEARNINGS.md into high-level files like AGENTS.md or SOUL.md. Be mindful that excessive logging can clutter the workspace, so prioritize quality over quantity. Always review the logs before starting a major task to ensure the agent is operating with the most recent insights.