Submit a skill
BENCH ENGINE · IDLENEXT CYCLE · —:—:—NEXT · —:—
Submit · CLAUDE.md

Bring your CLAUDE.md.

Drop your project context file. We classify it by project type (Next.js, Python data, backend API…) and rank it by how much it actually improves agent quality on real coding tasks for that type.

How we judge a CLAUDE.md

Same project, two runs of an agent: one with your CLAUDE.md mounted, one without. The bench engine measures the upliftin output quality. A CLAUDE.md that's clear, project-specific, and tightly scoped beats a generic one every time.

Tip: don't paste 800 lines of docs. The best CLAUDE.md files are 200-500 words of crisp project rules.

Heads up — sign in with GitHub to auto-claim a CLAUDE.md from your repo. Sign in ↗
Listing tier