Self Improving Agent 1.0.1
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Claude ('No, that's wrong...', 'Actually...'), (3) User requests a capability that doesn't exist, (4) An external API or tool fails, (5) Claude realizes its knowledge is outdated or incorrect, (6) A better approach is discovered for a recurring task. Also review learnings before major tasks.
Why this rating
Deterministic checks triggered by the tool capabilities and evidence.
- Locality: Local
Writes markdown logs under `.learnings/` in the project workspace.
- Data access: Sensitive
Learnings often include error output, commands, and project-specific context that may contain secrets or private code details.
- Action surface: Write
Appends to `.learnings/*.md` and may promote entries into files like `AGENTS.md` or `CLAUDE.md`.
Best practices
Follow these steps to reduce risk when using this skill.
- Redact secrets and credentials from error outputs before logging them to `.learnings/`.
- Keep `.learnings/` out of public repositories and shared artifacts unless reviewed and sanitized.
- Review entries before promoting to `AGENTS.md`/`CLAUDE.md` so incorrect or sensitive details don't become permanent.
Evidence links
Public sources backing the indicator assignments.
Always be careful when navigating away from the website.
Max-risk rule
If any capability reaches a higher level, the entire indicator level bumps up to keep ratings deterministic and easy to scan.