Self Improving Agent

Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...

This skill appears to do what it says (remind the agent to capture learnings and provide templates/helpers), but it explicitly instructs agents to copy full error contexts and source files into .learnings and to promote entries into shared workspace files or across sessions. Before installing or enabling hooks: 1) Review and sanitize the logging templates so they don't instruct agents to copy full source or paste logs containing secrets (remove guidance that says 'Full source of all included files'). 2) Prefer project-local .learnings directories over global ~/.openclaw/workspace so learnings don't get injected into unrelated sessions. 3) If enabling error-detector/activator hooks, run them with limited scope or add matchers (only trigger on specific commands or patterns) and confirm scripts are run from trusted working directories. 4) Audit any learnings before promoting to CLAUDE.md/AGENTS.md/SOUL.md or sending them cross-session. 5) If you need, ask the skill author to add redaction/scrubbing steps and an explicit opt-in for promotions that could expose code or credentials.

Type: OpenClaw Skill Name: mhbmissile-self-improving-agent Version: 1.0.0 The bundle provides a comprehensive framework for 'self-improvement' by instructing the agent to log errors, user corrections, and new insights into a `.learnings/` directory. It includes functional shell scripts (`activator.sh`, `error-detector.sh`, `extract-skill.sh`) and OpenClaw hooks (`handler.js`) that inject reminders into the agent's context to facilitate this feedback loop. While the skill encourages the agent to modify workspace files like `SOUL.md` and `AGENTS.md` to persist learnings, the logic is transparently aligned with its stated purpose and lacks any indicators of malicious intent, data exfiltration, or unauthorized execution.