self-improvement

To log errors and corrections for continuous improvement, you write entries into dedicated markdown files within the .learnings directory using a structured template. This creates a traceable learning log across agent interactions.

Capturing recurring patterns across OpenClaw sessions is best handled by using the skill's optional automation to detect and extract recurring behaviors. You can then promote these highlights into project memory files like AGENTS.md.

You promote widely applicable learnings into project memory by using the skill's optional automation to elevate highlights from the .learnings markdown files into configuration files like CLAUDE.md, AGENTS.md, or copilot-instructions.md.

Yes, you need frontmatter containing a name and description to track learning logs for agent workflows. This structured metadata is required alongside your dedicated markdown templates to ensure proper logging and knowledge reuse.

Yes, you can use this error-tracking workflow without external dependencies because the skill relies solely on structured markdown files and optional automation. It operates independently to log corrections and failed commands.

You should not use markdown files for logging agent corrections if you require automated database queries or complex data analytics, as this approach is strictly designed for text-based traceability and knowledge reuse.