self-improvement

You log agent errors and corrections to Markdown by capturing failed prompts, applied corrections, and discovered better approaches into centralized files, enabling cross-session memory and continuous agent behavior improvement.

Cross-session memory enables AI agents to retain learnings, errors, and corrections across different sessions by logging them to centralized Markdown files and applying OpenClaw hooks for knowledge sharing.

You promote a learning to project memory files like CLAUDE.md or AGENTS.md when a logged learning, error correction, or better approach becomes broadly applicable to the overall project context.

Integrate OpenClaw hooks by configuring them to trigger when prompts fail or corrections are made, automatically logging the learnings to Markdown files for cross-session knowledge sharing.

Optional scripts, references, and assets provide on-demand enhancements for capturing agent learnings, defining the discovery scope through frontmatter name and description while supporting the core logging mechanism.

No dependencies are required to log learnings and errors to Markdown files, as the skill operates independently using frontmatter for discovery scope and optional assets for enhancements.