I kept telling Claude "add rate limiting" and approving code without knowing what my users actually see when they hit the limit. A friendly message? A raw 429? Page just hangs?
VibeCheck asks you one question after your AI finishes a task. It reads your actual diff, compares it to what you asked for, and checks if you know what changed in your product. Not how the code works, but what your users experience.
It's a Claude Code stop hook (auto-triggers after every task) and also a standalone CLI that works with Cursor, Windsurf, Cline, Aider, OpenClaw, Nanoclaw or anything that uses git.
Single Rust binary, no dependencies, starts in under a millisecond. Skip every quiz if you want - no scores, no judgment.
frutigeraerosol•1h ago
VibeCheck asks you one question after your AI finishes a task. It reads your actual diff, compares it to what you asked for, and checks if you know what changed in your product. Not how the code works, but what your users experience.
It's a Claude Code stop hook (auto-triggers after every task) and also a standalone CLI that works with Cursor, Windsurf, Cline, Aider, OpenClaw, Nanoclaw or anything that uses git.
Single Rust binary, no dependencies, starts in under a millisecond. Skip every quiz if you want - no scores, no judgment.
https://github.com/akshan-main/vibe-check