clone-website · coco-research — Claude Code skill: what it does, how to install
clone-website replicates target web pages into a self-contained single-file HTML prototype without build tools. Handy when you need to inspect or recreate UI layouts quickly.
MIT ★ 328 updated 2026-10-29
npx skills add coco-research/coco --skill clone-website -a claude-codeWhat clone-website does
This skill uses Playwright to open target URLs and analyze the rendered page. It captures full-page and mobile screenshots, measures computed styles across elements, and extracts color usage, typography values, spacing, shadows, and corner radii. The reconnaissance phase also sweeps through buttons and tabs to capture interactive states.
The generated output is a single HTML file containing inline styles, inline scripts, and embedded SVG icons without external CDN links. You can run a complete clone by default, pass --extract to collect tokens and component specifications only, or supply --sections to clone specific blocks.
Playwright must be installed in the project before execution, because browser automation cannot run without it.
What it helps with
- generate a self-contained HTML prototype from a live URL
- extract computed CSS styles and design tokens via Playwright
- clone specific interface sections with the --sections flag
- bundle page structure and SVG icons into one dependency-free file
Example prompts for the agent
- Clone the interface at <url> into a single self-contained HTML file
- Extract design tokens and typography from <url> using the --extract flag
- Replicate only the navigation and table sections from <url> with --sections "header,table"
How to install clone-website
1Skills CLI (Vercel)
npx skills add coco-research/coco --skill clone-website -a claude-code2Manually: put the files in place
~/.claude/skills/clone-website/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/clone-website/
1Skills CLI (Vercel)
npx skills add coco-research/coco --skill clone-website -a codex2Manually: put the files in place
.agents/skills/clone-website/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add coco-research/coco --skill clone-website -a cursor2Manually: put the files in place
.cursor/skills/clone-website/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install coco-research/coco skills/clone-website2Manually: put the files in place
.github/skills/clone-website/SKILL.md1Skills CLI (Vercel)
npx skills add coco-research/coco --skill clone-website -a gemini-cli2Platform CLI
gemini skills install https://github.com/coco-research/coco --path skills/clone-websiteManually: put the files in place
.devin/skills/clone-website/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about clone-website
How to install clone-website in Claude Code?
Run npx skills add coco-research/coco --skill clone-website -a claude-code. You can also configure it manually at ~/.claude/skills/clone-website/SKILL.md.
What does clone-website output?
It outputs a standalone HTML file containing all styles, scripts, and embedded assets, ready to open in any web browser without build tooling.
What dependencies are required?
Playwright is mandatory. Check for it with npx playwright --version or install it into your project using npm i -D playwright.
Source text
SKILL.md · MIT
---
name: clone-website
description: Clone any website into a pixel-perfect single-file HTML prototype. Extracts design tokens, assets, CSS computed styles, interaction patterns, and content via Playwright. Outputs a self-contained HTML file with real data injection. Use when the user wants to clone, replicate, reverse-engineer, or create a pixel-perfect copy of any website or web app. Provide one or more target URLs as arguments.
argument-hint: "<url1> [<url2> ...]"
user-invocable: true
domain: design
---
# Clone Website --- Single-File HTML Prototype Builder
You are about to reverse-engineer **$ARGUMENTS** into a pixel-perfect single-file HTML prototype.
This is adapted from the [ai-website-cloner-template](https://github.com/JCodesMore/ai-website-cloner-template) approach but optimized for rapid prototyping workflows: **single self-contained HTML files** (no build system, no CDN dependencies, opens directly in a browser).
## Output Format
Unlike the original repo (Next.js + shadcn/ui), our output is:
- **One HTML file** with `<style>` + `<body>` + `<script>` sections
- **Zero external dependencies** --- all CSS inline, all JS inline, all assets base64-encoded or SVG inline
- **Real data injection** --- not lorem ipsum, but actual data from project Excel files, meeting notes, and brain DB
- **Project design system** when available --- use tokens from a `design-system.json` colocated with the project
## Pre-Flight
1. **Playwright is required.** Verify: `npx playwright --version`. If not installed, ask the user to run `npm i -D playwright`.
2. Parse `$ARGUMENTS` as one or more URLs. Validate each is accessible.
3. Create output directory: `{project}/Screenshots-clone/` for captured assets.
4. Determine clone mode:
- **Full clone** (default): pixel-perfect reproduction of the entire page
- **Design extraction only** (`--extract`): capture design tokens, screenshots, and component specs without building
- **Selective clone** (`--sections "header,table,sidebar"`): clone only named sections
## Phase 1: Reconnaissance (Playwright)
### 1.1 Screenshots
Use Playwright to capture:
```javascript
import { chromium } from 'playwright';
const browser = await chromium.launch();
const page = await browser.newPage({ viewport: { width: 1440, height: 900 } });
await page.goto(URL);
// Full page screenshot
await page.screenshot({ path: 'full-desktop.png', fullPage: true });
// Viewport screenshot
await page.screenshot({ path: 'viewport-desktop.png' });
// Mobile
await page.setViewportSize({ width: 390, height: 844 });
await page.screenshot({ path: 'viewport-mobile.png', fullPage: true });
```
### 1.2 Design Token Extraction
Run this in Playwright's `page.evaluate()` to extract the complete design system:
```javascript
const tokens = await page.evaluate(() => {
// Colors
const colorMap = new Map();
document.querySelectorAll('*').forEach(el => {
const cs = getComputedStyle(el);
['color','backgroundColor','borderColor','borderTopColor','borderBottomColor'].forEach(p => {
const v = cs[p];
if (v && v !== 'rgba(0, 0, 0, 0)' && v !== 'transparent') colorMap.set(v, (colorMap.get(v)||0)+1);
});
});
// Typography
const fontMap = new Map();
document.querySelectorAll('*').forEach(el => {
const cs = getComputedStyle(el);
const key = `${cs.fontFamily}|${cs.fontSize}|${cs.fontWeight}|${cs.lineHeight}`;
fontMap.set(key, (fontMap.get(key)||0)+1);
});
// Spacing
const spacingSet = new Set();
document.querySelectorAll('*').forEach(el => {
const cs = getComputedStyle(el);
['padding','margin','gap'].forEach(p => {
const v = cs[p]; if (v && v !== '0px') spacingSet.add(v);
});
});
// Shadows
const shadowSet = new Set();
document.querySelectorAll('*').forEach(el => {
const v = getComputedStyle(el).boxShadow;
if (v && v !== 'none') shadowSet.add(v);
});
// Radius
const radiusSet = new Set();
document.querySelectorAll('*').forEach(el => {
const v = getComputedStyle(el).borderRadius;
if (v && v !== '0px') radiusSet.add(v);
});
return {
colors: [...colorMap.entries()].sort((a,b) => b[1]-a[1]).slice(0,30),
typography: [...fontMap.entries()].sort((a,b) => b[1]-a[1]).slice(0,20),
spacing: [...spacingSet].sort(),
shadows: [...shadowSet],
radii: [...radiusSet].sort()
};
});
```
### 1.3 Component CSS Extraction (Deep)
For each major component/section, extract exact computed styles:
```javascript
// Run per component container
const componentCSS = await page.evaluate((selector) => {
const el = document.querySelector(selector);
if (!el) return null;
const props = [
'fontSize','fontWeight','fontFamily','lineHeight','letterSpacing','color',
'textTransform','textDecoration','backgroundColor','background',
'padding','paddingTop','paddingRight','paddingBottom','paddingLeft',
'margin','marginTop','marginRight','marginBottom','marginLeft',
'width','height','maxWidth','minWidth','display','flexDirection',
'justifyContent','alignItems','gap','gridTemplateColumns',
'borderRadius','border','boxShadow','overflow','position',
'top','right','bottom','left','zIndex','opacity','transform','transition'
];
function extract(element, depth) {
if (depth > 4) return null;
const cs = getComputedStyle(element);
const styles = {};
props.forEach(p => {
const v = cs[p];
if (v && v !== 'none' && v !== 'normal' && v !== 'auto' && v !== '0px' && v !== 'rgba(0, 0, 0, 0)')
styles[p] = v;
});
return {
tag: element.tagName.toLowerCase(),
classes: element.className?.toString().split(' ').slice(0,5).join(' '),
text: element.childNodes.length === 1 && element.childNodes[0].nodeType === 3
? element.textContent.trim().slice(0,200) : null,
styles,
children: [...element.children].slice(0,20).map(c => extract(c, depth+1)).filter(Boolean)
};
}
return extract(el, 0);
}, selector);
```
### 1.4 Interaction Sweep
Use Playwright to discover behaviors:
```javascript
// Scroll sweep
for (let y = 0; y < await page.evaluate(() => document.body.scrollHeight); y += 300) {
await page.evaluate(y => window.scrollTo(0, y), y);
await page.waitForTimeout(200);
// Check for sticky header changes, scroll-triggered animations, etc.
}
// Hover sweep --- hover over interactive elements
const interactiveEls = await page.$$('button, a, [role="tab"], .nav-item, .card');
for (const el of interactiveEls.slice(0, 30)) {
await el.hover();
await page.waitForTimeout(100);
}
// Click sweep --- test tabs, dropdowns, etc.
const tabs = await page.$$('[role="tab"], .tab-item, .nav-tab');
for (const tab of tabs) {
await tab.click();
await page.waitForTimeout(300);
}
```
### 1.5 Asset Extraction
```javascript
// Extract all SVG icons as inline code
const svgs = await page.evaluate(() =>
[...document.querySelectorAll('svg')].map(s => ({
html: s.outerHTML,
parent: s.parentElement?.className?.toString().split(' ')[0] || 'unknown',
width: s.getAttribute('width'),
height: s.getAttribute('height')
}))
);
// Extract all image URLs for download
const images = await page.evaluate(() =>
[...document.querySelectorAll('img')].map(i => ({
src: i.src, alt: i.alt, w: i.naturalWidth, h: i.naturalHeight
}))
);
```
## Phase 2: Foundation Build
Convert extracted tokens into CSS custom properties:
```css
:root {
/* Colors --- from extraction, mapped to semantic names */
--color-bg: {extracted};
--color-text: {extracted};
--color-text-muted: {extracted};
--color-accent: {extracted};
--color-border: {extracted};
/* ... */
/* Typography */
--font-family: {extracted};
--font-size-body: {extracted};
/* ... */
/* Shadows */
--shadow-sm: {extracted};
--shadow-md: {extracted};
/* ... */
}
```
## Phase 3: Parallel Section Building
For each major section of the page, dispatch a builder agent:
1. Write a spec file with exact CSS values + screenshot + content
2. Launch agent in a worktree (or as a subagent) with the spec
3. Each agent produces its section's HTML + CSS + JS
4. Orchestrator assembles all sections into one file
**Complexity budget**: If a section spec exceeds ~150 lines, split it.
## Phase 4: Assembly
Merge all sections into one self-contained HTML file:
1. Deduplicate CSS --- combine all section styles, remove duplicates
2. Order HTML --- sections in visual order (top to bottom)
3. Combine JS --- all event handlers and render functions in one `<script>` block
4. Base64-encode any images that couldn't be replaced with CSS/SVG
5. Verify: open in browser, compare against original screenshots
## Phase 5: Quality Check
Use Playwright to:
1. Open the built HTML file
2. Take screenshots at the same viewpoints as Phase 1
3. Compare side-by-side (save both for manual review)
4. Test all interactions (clicks, hovers, tabs)
5. Report any visual differences
## Project-Specific Mode
When a `design-system.json` exists in the project:
- Load tokens (colors, typography, spacing, shadows, radii) from that file
- Apply them as CSS custom properties in the output
- Inject any real data referenced in the project (Excel, JSON, brain DB)
## Reference
- Adapted from [JCodesMore/ai-website-cloner-template](https://github.com/JCodesMore/ai-website-cloner-template)
© coco-research · MIT · coco-research/coco
Similar skills
All skills for Frontend and UI →clone-website by other authors comes first.