arvindrk-extract-design-system

extract-design-system — Extract Design Tokens from Any Website

Reverse-engineer design tokens (colors, typography, spacing, border radius, shadows) from any public website and generate starter JSON and CSS custom properties for your local project. Available as an AI agent skill (Claude, Cursor, Codex) and standalone CLI.

Skills follow Anthropic's Agent Skills overview and are installable through the skills.sh ecosystem.

This repository is skills-first. It publishes an installable extract-design-system skill, and the bundled CLI is the executable workflow the agent runs under the hood.

What It Extracts

Token category	Examples
Colors	brand palette, backgrounds, text, borders
Typography	font families, sizes, weights, line heights
Spacing	padding and margin scale
Border radius	button, card, and pill radii
Shadows	box-shadow values across elevation levels

Outputs a W3C-compatible tokens.json and a tokens.css file of CSS custom properties, ready to drop into any project.

Quick Start

Install the skill:

npx skills add arvindrk/extract-design-system

Run it from a supported coding agent with a public website:

Extract the design system from https://stripe.com and generate starter token files for this project.

Expected starter outputs:

• .extract-design-system/raw.json
• .extract-design-system/normalized.json
• design-system/tokens.json
• design-system/tokens.css

About This Repository

This repository contains a focused agent skill for reverse-engineering a starter design system from a public website.

The skill is designed to give coding agents a repeatable workflow for:

• analyzing a public website's colors, typography, spacing, radius, and shadows
• normalizing the extracted output into a stable structure
• generating project-local starter token files
• summarizing findings before broader styling or app changes

Available Skill

extract-design-system

Extract a starter design system from a public website and turn it into local token files.

Use when:

• You want an agent to analyze a public website's visual primitives
• You want starter design tokens for an existing project
• You want a repeatable workflow instead of ad hoc prompting
• You want the agent to summarize findings before applying broader changes

What it produces:

• .extract-design-system/raw.json
• .extract-design-system/normalized.json
• design-system/tokens.json
• design-system/tokens.css

Installation

Install the skill from GitHub with the skills CLI:

npx skills add arvindrk/extract-design-system

You can also browse the skill on skills.sh.

Usage

Once installed, the skill is available to supported coding agents. The agent should use it when a request matches the extraction workflow.

Example prompts:

Extract the design system from https://stripe.com and generate starter token files for this project.

Analyze https://linear.app and summarize the design primitives before generating local tokens.

Extract colors, typography, spacing, and radius tokens from https://vercel.com and save them for this codebase.

Review https://tailwindcss.com, extract the most likely design primitives, and generate starter token files only if the results look usable.

How The Skill Works

The skill is designed to:

• confirm that the target site is public and reachable
• set expectations about scope and limitations
• run the extraction workflow through the bundled CLI
• summarize what was found in the normalized output
• generate starter token files for the current project
• ask before modifying existing app code, styles, or configuration

Scope And Limitations

• Public websites only
• Single-page extraction workflow
• Starter tokens, not a full component library
• No framework config patching or automatic app rewrites
• Dynamic, protected, or highly script-driven sites may yield incomplete output
• Extraction is useful for initialization, not pixel-perfect reproduction

Security Considerations