The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes several Bash scripts (activator.sh, error-detector.sh, extract-skill.sh) designed to assist with learning capture and skill extraction. These scripts are intended for local execution by the agent or the underlying platform hook system.
[COMMAND_EXECUTION]: The extract-skill.sh utility automates the creation of new skill templates. It implements robust security checks, including regex-based validation for skill names and strict path traversal prevention that blocks absolute paths and '..' segments.
[EXTERNAL_DOWNLOADS]: Documentation references installation via a GitHub repository (github.com/peterskoett/self-improving-agent) and a package manager (clawdhub). These are standard installation methods for the mentioned agent platforms.
[SAFE]: The skill utilizes OpenClaw hooks (handler.js, handler.ts) to inject virtual markdown files into the agent's bootstrap process. This is a standard and documented extension pattern for providing behavioral reminders and context-aware guidance.
[SAFE]: Instructions for 'promoting' learnings to project memory files like CLAUDE.md and AGENTS.md are documented as a manual or agent-driven workflow to improve long-term session performance.

self-improvement