self-improvement

You can capture and log coding learnings by writing them to .learnings/LEARNINGS.md and errors to .learnings/ERRORS.md. This skill uses activator and error-detector hooks to automatically record corrections encountered during AI workflow development.

Tracking recurring errors in OpenClaw environments is handled by an error-detector hook that logs issues to .learnings/ERRORS.md. A scripted extraction path then converts these recurring insights into reusable skills for future sessions.

Promoting useful learnings to shared project memory involves moving recurring insights into CLAUDE.md, AGENTS.md, and copilot-instructions files. This structured promotion pathway ensures broadly useful corrections are accessible across the project.

Yes, you can log feature requests during development by writing them to .learnings/FEATURE_REQUESTS.md. This runs parallel to the error and learnings tracking, keeping all continuous improvement data organized in one directory.

No, this continuous improvement workflow requires zero external dependencies. It operates independently using internal scripts, references, and assets to manage the logging, extraction, and promotion of development learnings.

Automated error tracking hooks are not suitable when your workflow lacks a persistent file system for the .learnings directory. The structured extraction and promotion process requires stable local storage to manage logs and shared guidance files.