Part of an experiment

This is the Light agentic with review build — agents implement freely, with a review gate that can send work back before it commits. See the video, blog post, and source code for the full study.

Blog post Video Source

CLAUDE.md Scorer

Evaluate your CLAUDE.md file across 6 dimensions

📋

Drop your CLAUDE.md here or click to upload

Or try a sample

Scoring your file...

Score Report

Overall: / 10

All Findings

Suggestions

Live Evaluation

Test how well Claude follows your CLAUDE.md rules by sending 3 coding tasks.

⚠ We will not save your key. It is sent directly to the Anthropic API and discarded.

Running evaluation (this may take a minute)...