| name | website-extractor |
| description | Extract all content and design information from a website to enable a complete rewrite. Use when the user wants to rewrite, redesign, or create a new version of an existing website. Captures page content, brand identity (colors, fonts, tone), navigation structure, section-by-section copy, CTAs, and visual style. Triggers on "extract this website", "rewrite this site", "create a new version of", "redesign based on", or any request that involves analyzing an existing site before producing new content or code. |
Website Extractor
Extract all information needed to rewrite or redesign a website. Output a structured document that a rewrite agent can use directly.
Workflow
Step 1 — Open the site and screenshot
agent-browser open <url>
agent-browser screenshot --full site-full.png
Read the screenshot to get an overall visual impression: layout style, color palette, density, imagery approach.
Step 2 — Get the full accessibility snapshot
agent-browser snapshot
This gives the full content tree: all headings, paragraphs, links, buttons, and labels in document order. Use this as the primary source for text content.
Step 3 — Get the HTML for design signals
agent-browser get html
Scan the HTML for:
- Tailwind classes: color classes like
bg-indigo-600, text-gray-900, border-blue-500 directly reveal the palette
- Custom CSS variables: look for
var(--color-*) patterns
- Font references:
font-family declarations or Google Fonts <link> tags
- Framework clues: Next.js, Webflow, WordPress, etc.
Step 4 — Navigate additional pages (if multi-page)
For each key page (About, Pricing, Features, etc.) visible in the nav:
agent-browser open <page-url>
agent-browser snapshot
Extract content from each page. Note which pages exist.
Step 5 — Produce the extraction document
Write a structured markdown document following references/extraction-format.md.
Save it as site-extraction.md in the current directory (or wherever the user specifies).
Tips
- Prioritize the snapshot over raw HTML for content — it's already cleaned up and readable
- For colors: Tailwind class names are exact hex values; look them up if needed. For custom CSS, note the variable names and any hex values found in
<style> tags
- For tone: Read the hero and feature copy. Is it formal/casual, technical/approachable, bold/understated?
- Don't guess colors from screenshots — always extract from code; screenshots compress colors
- If the site is SPA/JS-rendered and snapshot is empty, scroll first:
agent-browser scroll down 500 then re-snapshot
- If agent-browser isn't installed:
npm install -g agent-browser && agent-browser install
Output
The extraction document is the deliverable. Tell the user where it was saved and offer to immediately pass it to a rewrite tool or describe what you found.