jaredrhod/barehands ? reverse-engineered prompt

Build me a webcam based screen interface where I can move notes, images, and 3D models around with hand gestures in Chrome. I want a simple local page that opens a camera view, tracks my hands, and shows floating glass like cards I can tap, grab, stretch, throw, and expand into an exploded view with two fingers. It should feel like a playful board for my AI, with a ring that can reflect whether the assistant is thinking, listening, idle, or speaking, and a way for the assistant to put things on the board or read what is already there.

Please make it easy to run locally with a tiny server, and let me point it at a folder of markdown notes and a media folder for images and models. Keep it safe so only files from the allowed media folder can appear. If you need current docs for anything, look them up online.