Released an agentic computer vision framework on npm. Repo: https://github.com/vfrog-ai/croak It works inside Claude Code and walks you step-by-step through building an object detection model — from raw images to deployed endpoint. Try it out. Would love feedback on what's useful (and what's not).