kris77z/web-experience-cloner

This project is a skill you can give to an AI agent, not an app you run yourself. Its job is to help that agent reverse engineer, copy, take offline, and rewrite complicated websites. It is aimed at tricky sites built with creative animation tools such as WebGL, Three.js, GSAP, Rive, or gaussian splatting, as well as sites exported from design platforms like Framer, Webflow, and Readymag, and app style pages built with frameworks such as React Router, Remix, Oxygen, Next.js, or Nuxt. The skill lives in a single folder, and its main file is called SKILL.md. Any AI agent that understands this kind of skill folder format can read it and follow its instructions. You can add it with a command line tool called skills, or simply copy the folder into your agent's own skills directory, including a Codex specific path if you use that agent. The workflow the skill follows has six steps in order: classify the site, gather evidence about how it was built, lock down that understanding, trace how it actually behaves, then either mirror or replay it, and finally verify the result. The idea is to figure out what kind of site it is before trying to copy it. The project defines five levels of output an agent can produce, from simplest to most involved. The lightest is just classifying the site and sketching a plan. Next is documenting how a specific animation works. After that comes a full local copy of the site that reuses its original code. Beyond that is making that copy work fully offline. The most advanced level is rewriting a single visual effect completely from scratch. Before publishing changes, there is a script that checks the skill folder for structural problems. The code and documentation are released under the MIT license, but the README is clear that copying a live site with this tool still means respecting that site's copyright, terms of service, and local law, and copies should stay for private study rather than being redistributed.