self-improvement

What This Skill Does

The Self-Improvement skill acts as a long-term memory and error-tracking engine for the OpenClaw AI agent. By institutionalizing the logging of failures, user corrections, and feature gaps, it transforms ephemeral interaction sessions into a persistent knowledge base. This skill automates the creation of structured markdown logs in a dedicated .learnings/ directory, ensuring that the agent can review historical mistakes and successful patterns before tackling new tasks. It serves as a continuous feedback loop that bridges the gap between ad-hoc debugging and systematic project growth.

Installation

To integrate this skill, navigate to your project root and initialize the storage directory:

mkdir -p .learnings

Once the directory is created, ensure your agent has write permissions for the path. You can then populate the directory with the standard LEARNINGS.md, ERRORS.md, and FEATURE_REQUESTS.md templates provided in the documentation to ensure consistent metadata tracking across your development lifecycle.

Use Cases

This skill is essential for teams aiming to reduce technical debt and minimize recurring agent errors. Use it whenever you hit a wall: when an API call fails, the agent generates a hallucinated syntax, or you find yourself manually correcting the agent's logic. It is particularly powerful for complex, multi-stage deployments where tracking the 'why' behind an error is as critical as the fix itself. By tagging entries with categories like 'knowledge_gap' or 'best_practice', the agent can later query its own history to select the most efficient approach for current tasks.

Example Prompts

1. "The last command failed with a timeout error; please log this in the ERRORS.md file with the relevant context so we can troubleshoot it later."
2. "Actually, that approach wasn't quite right because of the dependency constraint; let's capture this correction as a 'knowledge_gap' learning entry."
3. "I wish I could automate the documentation generation process; can you add this as a new feature request in the tracker?"

Tips & Limitations

Maintain discipline by reviewing the .learnings/ directory regularly. The utility of this skill depends entirely on the quality of the metadata; be descriptive in your 'Summary' and 'Details' sections. Avoid cluttering the files with trivial, one-off events that don't offer value for future iterations. Finally, remember that this skill creates files in your local repository, so ensure these files are included in your .gitignore if you do not wish to version control the agent's internal thought logs, or conversely, commit them if you want the improvement history to persist across developer machines.