AI Skill MartBrowse skillsPromptsProgramsLearnWhat is a Skill?
 
← Back to list

Darwin Skill 2.0 (Autonomous SKILL.md Optimizer)

A meta-skill that scores SKILL.md files on a 9-dimension rubric and improves them through an edit → test → human-approval → keep-or-revert loop backed by git.

UtilitiesAdvanced★ 6,846⑂ 710AI score 9/10Last updated: Sep 18, 2026

What it does

  • Scores any SKILL.md out of 100 across nine dimensions: frontmatter quality, workflow clarity, failure-mode encoding, checkpoint design, actionable specificity, resource wiring, overall architecture, measured performance, and an anti-pattern blacklist.
  • Creates an auto-optimize/YYYYMMDD-HHMM branch and runs a hill-climbing loop that changes only one dimension per round.
  • Decides keep vs. revert with paired same-judge comparisons and odd-N majority voting, instead of absolute score deltas (which carry ±8 points of judge noise).
  • Rolls back with git revert rather than reset --hard, and logs every experiment to results.tsv.
  • Designs 2–3 test prompts per skill and runs sub-agents with and without the skill to compare real output quality.
  • Generates a PNG result card from three HTML themes (Swiss / Terminal / Newspaper).
  • Runs a runtime-neutrality red-flag grep so the skill isn't rejected by non-Claude-Code agents.

Who it's for

  • Skill authors who want an objective quality gate before publishing.
  • Maintainers of a growing skill library that needs triage and cleanup.
  • Anyone who prefers experiment-driven prompt improvement with logs and rollback.

Example uses

  1. "Optimize all my skills" → scans every SKILL.md, prints a baseline scoreboard, waits for your confirmation, then fixes the weakest first.
  2. "Just evaluate skill X" → designs test prompts and scores the baseline only, without editing any file.
  3. "How good is my skill?" → diagnoses the largest weighted gap, stops at a checkpoint, and on approval applies one improvement round with a git diff and result card.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/alchaincyf/darwin-skill/HEAD/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on AI Skill Mart.
Copy the SKILL.m folder from the GitHub repo alchaincyf/darwin-skill into my ~/.claude/skills/darwin-skill/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/alchaincyf/darwin-skill.git ~/.claude/skills/darwin-skill

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal and make sure the skills folder exists: mkdir -p ~/.claude/skills
  2. Clone the repository into it: git clone https://github.com/alchaincyf/darwin-skill.git ~/.claude/skills/darwin-skill
  3. Verify that SKILL.md, plus the templates/ and scripts/ folders, are present.
  4. For PNG result cards, install Node.js and then run: npx playwright install chromium
  5. The skills you want to optimize must live inside a git repository for the rollback ratchet to work; run git init first if needed.
  6. Restart Claude Code and type "use darwin to review my skill" (or the Chinese trigger "优化skill") to activate it.
  7. On your first run, ask for "evaluation only" to inspect the scoreboard before letting it edit files.
View source on GitHub ↗License: MIT