AI-powered code analysis and auto-fixing CLI tool
Reproducible adversarial eval for AI agents. Deterministic scoring, no model judge. Vibefixing publishes the eval; we don't compete on the leaderboard.