darwin-skill

darwin-skill

darwin.skill —— Automatically optimize your Agent Skills by evaluating, improving, testing, retaining, or rolling back changes.

Agent SkillDevOpen sourceautoresearchskill-optimizationclaud-code
Type
Agent Skill
Open source
Yes
GitHub Stars
★ 4.3k
Source
skill-github

Overview

darwin.skill is an autonomous skill optimization system inspired by autoresearch, specifically designed for Claude Code. It helps continuously improve Agent Skills through a process of evaluation, enhancement, testing, and either retaining or rolling back changes. Drawing on research from Microsoft Research's SkillLens and SkillOpt, it provides a structured framework with verifiable outcomes. Ideal for developers managing multiple Agent Skills who seek ongoing performance improvements.

Capabilities

  • Automatic evaluation of Agent Skills
  • Generation of specific improvement proposals
  • Multi-dimensional scoring system
  • Independent sub-agent scoring
  • Ratcheting mechanism to preserve improvements

Use cases

Optimizing existing Agent SkillsEnhancing the practical effectiveness of Agent SkillsAutomated maintenance of large numbers of Agent Skills

Setup

Requires: API KeyNode 环境
npx skills add alchaincyf/darwin-skill

This information was compiled by AI from public sources and may contain inaccuracies — please refer to the source.

FAQ

After installation, say "optimize all skills" or "optimize a specific skill" in any supported Agent tool that handles SKILL.md.

After installation, say "optimize all skills" or "optimize a specific skill" in any supported Agent tool that handles SKILL.md.

Which platforms does darwin.skill support?

Supports any Agent tool compatible with the SKILL.md format.

Related skills