Clone any website with AI agents that don't cut corners
A new open-source agent skill lets developers recreate any website with pixel-perfect accuracy using /clone-website. Instead of guessing layouts in one massive prompt, it breaks components into strict specs under 150 lines and builds them in parallel git worktrees before running visual diffs. It basically fixes the habit of AI faking CSS spacing and fonts.
clone-website is an open-source agent skill designed to recreate any webpage with pixel-perfect precision. Run /clone-website alongside a URL, and agents will extract design tokens, pull assets, and rebuild the interface from scratch.
How it stops agents from cutting corners
Most AI coding tools fail on full-page clones because the models start hallucinating CSS values, approximating padding, and guessing font sizes. This workflow sets hard engineering boundaries so the agent cannot cheat.
First, it enforces a strict 150-line limit on component specifications. If a block is too complicated, it must be split up. Builders only receive isolated components paired with real computed styles directly from getComputedStyle(), leaving zero room for guesswork.
Before writing code, the system writes a strict spec containing screenshots, real text, asset paths, and element states. It also scrolls the page first to verify whether an animation is scroll-driven or click-driven, avoiding massive rewrites later. Separate builder agents then work in parallel within isolated git worktrees before merging and running visual diffs against the live site.
Why it matters
Extracting web layouts is not new, but fixing the messy code that AI usually produces usually takes more time than building it manually. This project shows that if you restrict agents to tiny tasks, feed them grounded browser data, and verify output with visual diffs, AI can deliver production-grade UI replication. It is a brilliant blueprint for any developer building autonomous coding pipelines.
Sources (1)
- github.com(opens in a new tab)Originalgithub.com