What problem does it solve? AI-generated website clone analyses often contain fabricated code that breaks when run. This Skill enforces a real-source-first methodology for reproducing websites, from static pages to WebGL-heavy interactive demos, with executable probes and evidence grading instead of hallucinated code. ## Core Features & Use Cases - Six-step decision tree: Search GitHub for real source first, then probe the live site with Playwright scripts (recon, route crawling, network capture, interaction probing, sourcemap hunting) before choosing a clone path. - WebGL/Canvas reverse engineering: Dissect shader-based frontends line by line with SOURCE/PARTIAL/GUESS evidence grading and a baseline-first replay gate before refactoring. - Static-site 1:1 mirroring: Full-scroll capture and mirror every deployed asset (including runtime-fetched .sog/.buf/.wasm files) for Astro/Vite SSG/Hugo sites. - Design DNA extraction: Distill a site's visual identity into a versionable design-dna.json for visual-clone or content-swap modes. - Use Case: Clone a Three.js marketing site by running recon-site.mjs and mirror-site.mjs, then generate a CLONE_REPORT.md with pixel-diff scores comparing original and clone. ## Quick Start Ask the AI to clone a website by providing its URL, for example: clone this site for me: https://example.com.
