Self Improving Skills — Design skill for Claude Code

Self-Improving Skills

**English** | [한국어](README.ko.md)

**Hermes Agent-style self-improvement for Claude Code, Claude Cowork, Codex, and ChatGPT work.**

It turns hard-won workflow lessons into reusable `SKILL.md` files, validates skill edits, curates stale knowledge, and (since v0.13.0) distills in a **detached background worker** — your visible turn ends normally while a headless session captures the technique.

Claude Code already has hooks, subagents, slash commands, and skills. This plugin wires those primitives into a closed learning loop inspired by [Nous Research Hermes Agent](https://github.com/NousResearch/hermes-agent):

complex task → distill what worked → save/patch a skill → rediscover next session

Variants

One repo, four ports of the same closed learning loop:

Plugin	Environment
claude-code-self-improving-skills	Claude Code CLI (this README's main subject)
claude-cowork-self-improving-skills	Claude Cowork (cloud container) — persists skills via claude.ai '스킬 저장'
chatgpt-codex-self-improving-skills	OpenAI Codex CLI (hooks + MCP skill manager)
chatgpt-work-self-improving-skills	ChatGPT desktop Work (skills-only package in the shared plugins/ directory)

Install one per environment — do not install two variants into the same environment (duplicate hooks/nudges).

Why this exists

Most coding agents can solve a hard problem once. Fewer can reliably remember the reusable part of that work and apply it later.

Hermes Agent has first-class procedural memory through skills and a curator loop. This project ports that idea into Claude Code as a plugin:

• Detect complex work using transcript/tool-call signals.
• Nudge the agent at the right time to distill the reusable technique.
• Write or patch ~/.claude/skills//SKILL.md through a dedicated subagent.
• Validate and roll back malformed skill edits automatically.
• Track usage so stale generated skills ca