Self Improving Agent

A universal learning system that watches every skill execution and automatically improves the codebase based on what went wrong or right. It maintains three memory types: semantic patterns for reusable rules, episodic memories of specific debugging sessions or PRD failures, and working memory for current context. When you hit an error or complete a task, it extracts the experience, abstracts it into patterns, and updates relevant skills with new checklist items, anti-patterns, or quality rules. Built on 2025 lifelong learning research with hooks that auto-trigger on skill events. Essentially turns your Claude instance into something that gets better at coding, debugging, and planning the more you use it.

install copies · 90 days. Times the install command was copied on this site in the last 90 days, bots filtered.: 6
skills.sh installs: 31.9K
GitHub stars: 66
last updated: Jul 14, 2026

View on GitHub

Install

$ npx -y skills add charon-fan/agent-playbook --skill self-improving-agent --agent claude-code

Installs into .claude/skills of the current project.

Files

Self-Improving Agent

Turn evidence from completed work into a small, auditable behavior change. The default result is a candidate or no change—not an automatic rewrite of skills.

Use This Skill When

A tool or workflow failed in a way that may recur.
The user corrected an assumption, requirement, or operating rule.
The same workaround appeared more than once.
A focused test proved a better reusable method.
The user asks to review or consolidate learning candidates.

Do not use it for routine session summaries, raw transcript storage, speculative ideas without evidence, or project facts that belong in project documentation.