Gauntlet Loop

Gauntlet Loop is an AI agent skill for people who want generated websites, writing or software judged against a specific example. It writes a prompt that asks an agent to build, compare and revise work against a reference you choose.

• Concrete references. The skill proposes named pages, articles or repositories and checks that the agent can access them and make a meaningful comparison.
• Independent critics. The prompt divides the task into smaller pieces and assigns separate builders and critics, with fresh context for each critic.
• Blind comparison. Critics inspect the actual output beside the reference with identifying labels removed, then choose which is better rather than assign a score.
• Repeated revision. The prompt directs the agent to keep revising until the critic prefers its work or you stop the run, without a fixed round limit.

It packages Matt Shumer's gauntlet loop technique as a reusable skill. It works with Claude Code and adapts its instructions for other agents that can run parallel builders and critics. The skill supplies prompts; the chosen agent performs the work and needs access to the reference. It's licensed under Creative Commons Attribution 4.0 International (CC BY 4.0), which requires attribution.