promptwhisper/clone-any-website-skill

This project is a skill for the Codex coding assistant that guides it through rebuilding a public website as a clean, maintainable project instead of just copying the original files. Rather than performing the work itself, the repository provides structured instructions and reference documents that Codex reads and follows step by step when a user asks it to recreate a site. The workflow covers deciding what kind of technology stack the target site uses, whether it is a normal DOM based page, a Three.js 3D scene, a Pixi.js or Canvas2D animation, or a video heavy landing page. It then covers capturing how the site looks and behaves across different browser states and screen sizes, figuring out how users are meant to interact with it through scrolling, clicking, hovering, or time based animation, and writing that understanding down as a specification before rebuilding the page. It also covers mirroring static assets like images locally, frame aligned reconstruction of MP4, WebM, or HLS video content, and looking at a site's existing code bundle only as a reference for how it behaves, not to copy it directly. After rebuilding, the skill checks the result with screenshot comparisons, interaction testing, network behavior checks, and build and deployment validation. The README is explicit that this workflow is meant for public targets only, and that it should not be used to bypass authentication, payment systems, access controls, or other technical restrictions. The goal is to reimplement a site's visible behavior as original source code, not to redistribute or repackage the original compiled application. To install it, you clone the repository and create a symbolic link into your Codex skills folder, then restart Codex. You can trigger it with an explicit command naming the skill, or with natural requests like asking it to clone an interactive Three.js page or rebuild a video landing page frame by frame. The repository includes detailed reference documents for each technology area, research templates, and a Python script that compares screenshots and reports pixel level differences. It is released under the MIT license.