What makes a repo ready for AI coding agents
Every score on this site comes from 16 checks on the files in a repo. Each one has its own page: what it looks for, why agents need it, how to pass it, and how many of the 371 repos on the leaderboard do. The methodology covers how the checks are weighted into a score.
- AGENTS.md / CLAUDE.md50% pass
Presence of an agent-oriented instructions file, with substantive content.
- Cursor rules (.cursor/rules)4% pass
Cursor's canonical instruction surface —
.cursor/rules/*.mdc(modern) or.cursorrules(legacy). - GEMINI.md1% pass
Gemini CLI's canonical hierarchical instructions file — read at every prompt.
- .openhands/setup.sh0.5% pass
OpenHands runs
.openhands/setup.shat session start to bootstrap the repo's dev environment. - .aider.conf.yml0.3% pass
Aider reads
.aider.conf.yml(or.yaml) for repo-level config — model, lint command, test command. - README96% pass
Non-trivial README so the agent can learn the project quickly.
- Test suite52% pass
Detectable tests — agents rely on feedback loops.
- CI configuration95% pass
Defined pipeline the agent can reason about / emulate locally.
- Linter / formatter config69% pass
Agents get immediate feedback on style rather than ambiguous drift.
- Dependency manifest92% pass
Machine-readable dependency list so the agent can reproduce the env.
- Reproducible dev env21% pass
One-command setup the agent can run (Makefile / devcontainer / Nix / Docker).
- Type configuration67% pass
Static types help agents reason about call sites without running code.
- License file91% pass
Clarity on what an agent is allowed to do with the code.
- CONTRIBUTING guide87% pass
Explicit contribution workflow an agent can follow.
- Pre-commit / git hooks31% pass
Catches problems locally before the agent wastes a CI cycle.
- Manageable size16% pass
Very large repos strain an agent's context window.