# alchaincyf/darwin-skill

达尔文.skill —— 一个让你的Skill无限进化的系统：评估→改进→测试→保留或回滚 | Autoresearch-inspired autonomous skill optimization for Claude Code. Evaluate, improve, test, keep or revert.

Repository: https://github.com/alchaincyf/darwin-skill
Canonical: https://ross.abutalabs.com/products/darwin-skill
Language: HTML
License: MIT
License Family: permissive
Last push: 2026-08-25T03:01:10+00:00

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 35, longevity 10
- inputs: {"age_days": 142, "days_push": 8, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5754, forks 612 (observed 2026-08-28T04:09:29.066892+00:00)

## What it is
Darwin.skill is an agent skill for Claude Code that autonomously optimizes other SKILL.md agent skills through an evaluate-improve-test-keep-or-revert loop, inspired by Karpathy's autoresearch and aligned with Microsoft Research's SkillOpt validation-gated framework. It scores skills across nine rubric dimensions, uses independent multi-judge review, and enforces human-in-the-loop checkpoints and regression gates so only genuinely improved changes are kept.

## Use cases
- automatically improve my claude code skills
- evaluate and score agent skill quality
- optimize skil.md files like training a model
- run regression tests on agent skills before keeping changes
- bulk maintain a large collection of agent skills
- set up an autonomous skill optimization loop
- audit agent skills for failure modes and risky actions

## When to choose
- you maintain many Claude Code / SKILL.md-format agent skills and need systematic quality improvement
- you want evaluation-gated optimization with human checkpoints rather than fully automatic rewrites
- you want rubric-based scoring of both structure and real-world effectiveness of skills

## When to avoid
- you don't use Claude Code or any SKILL.md-compatible agent tool
- you need fully autonomous optimization with no human-in-the-loop pauses
- you only have one or two simple skills where manual editing is cheaper

## Facets
- artifact type: plugin
- maturity: active
- function: agent-framework, prompt-engineering, testing, benchmarking, workflow-automation
- domain: developer-tools, artificial-intelligence
- platform: cli, cross-platform
- tags: claude-code, agent-skills, skill-optimization, autoresearch, skillopt, evaluation-loop, llm-tools, ai-agents, automation, nodejs

## Member repositories
- alchaincyf/darwin-skill (main) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:29.066892+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:53:16.965665+00:00, confidence not recorded.
  - readme: https://github.com/alchaincyf/darwin-skill (fetched 2026-08-28T04:09:29.066892+00:00, sha 339e75c4a285)
- Data as of 2026-08-30T08:39:29.467469+00:00.