skills/arustydev/ai/design-tokens-extraction

design-tokens-extraction

SKILL.md

Design Tokens Extraction

Extract and normalize design tokens from design files into standardized formats.

Overview

This skill enables extraction of design tokens from Figma, Sketch, and Penpot design files using their respective MCP servers. Extracted tokens are normalized to the W3C Design Tokens Community Group format for platform-agnostic storage and conversion.

This skill covers:

  • Connecting to design files via MCP servers (Figma, Sketch, Penpot)
  • Extracting color, typography, spacing, and shadow tokens
  • Normalizing extracted data to W3C DTCG format
  • Converting to platform-specific formats (Swift, CSS, Tailwind)

This skill does NOT cover:

  • Design file creation or modification
  • Visual design decisions
  • Component code generation (see design-to-swiftui skill)

Quick Reference

MCP Server Tools

Server Tool Purpose
figma get_file Read Figma file structure
figma get_styles Extract published styles
figma get_local_variables Extract variables/tokens
sketch-context get_document Read Sketch document
sketch-context get_shared_styles Extract shared styles
sketch-context get_color_assets Extract color variables
penpot get_file Read Penpot file
penpot get_components List design components

Token Categories

Category Properties Extracted
Color hex, rgb, hsl, opacity, description
Typography family, size, weight, lineHeight, letterSpacing
Spacing value (px, rem), description
Shadow color, offsetX, offsetY, blur, spread
Border Radius value (px, rem)
Duration value (ms)

Workflow: Extract Tokens from Design File

Step 1: Connect to Design Source

Determine the design source and use the appropriate MCP server:

Figma:

Use figma MCP server with get_file tool
Input: file_key from Figma URL (e.g., "abc123" from figma.com/file/abc123/...)

Sketch:

Use sketch-context MCP server with get_document tool
Input: file path to .sketch file

Penpot:

Use penpot MCP server with get_file tool
Input: project_id and file_id from Penpot URL

Step 2: Extract Raw Design Data

Query the MCP server for design tokens:

Colors:

  1. Get published color styles (Figma: get_styles, Sketch: get_shared_styles)
  2. Get color variables (Figma: get_local_variables, Sketch: get_color_assets)
  3. Extract from component fills if no styles defined

Typography:

  1. Get text styles from design system
  2. Extract font family, size, weight, line height, letter spacing
  3. Map font weights to numeric values (400, 500, 600, etc.)

Spacing:

  1. Look for spacing tokens in variables/styles
  2. Extract from auto-layout settings if available
  3. Check for spacing scale documentation

Shadows:

  1. Get effect styles (drop shadows, inner shadows)
  2. Extract color, offset, blur, spread values

Step 3: Normalize to W3C Format

Transform extracted data to W3C Design Tokens structure:

{
  "$schema": "https://design-tokens.github.io/community-group/format/",
  "color": {
    "<category>": {
      "<name>": {
        "$value": "#hexcode",
        "$type": "color",
        "$description": "From design file"
      }
    }
  },
  "typography": {
    "fontFamily": { ... },
    "fontSize": { ... },
    "fontWeight": { ... }
  },
  "spacing": { ... },
  "shadow": { ... }
}

Step 4: Output in Requested Format

Use the appropriate output style:

  • design-token-json - W3C DTCG format (default)
  • design-token-swift - Swift extensions
  • design-token-css - CSS custom properties
  • design-token-tailwind - Tailwind config

Extraction Patterns

Figma Color Extraction

When reading from Figma:

  1. Published Styles (preferred):

    • Use get_styles with style_type: "FILL"
    • Returns named color styles with descriptions
  2. Variables (Figma variables):

    • Use get_local_variables
    • Filter by resolvedType: "COLOR"
    • Group by collection name
  3. Component Fills (fallback):

    • Traverse component nodes
    • Extract unique fill colors
    • Auto-generate names from component path

Sketch Color Extraction

When reading from Sketch:

  1. Shared Styles:

    • Use get_shared_styles with style_type: "layer"
    • Extract fill colors from style attributes
  2. Color Assets:

    • Use get_color_assets
    • Returns swatch library colors
  3. Document Colors:

    • Access document-level color palette
    • Often contains brand colors

Penpot Color Extraction

When reading from Penpot:

  1. Color Library:

    • Access file's color library
    • Returns named colors with metadata
  2. Component Colors:

    • Extract from component definitions
    • Track color references

Normalization Rules

Color Normalization

Source Format Target Format
rgba(r, g, b, a) #RRGGBBAA
rgb(r, g, b) #RRGGBB
hsl(h, s%, l%) #RRGGBB (converted)
{ r, g, b } (0-1) #RRGGBB (scaled)
{ r, g, b } (0-255) #RRGGBB

Typography Normalization

Property Source Variations Target
Font Size 12px, 0.75rem, 12 "0.75rem" or 12
Font Weight "bold", 700, "semibold" 700 (numeric)
Line Height 1.5, 150%, 24px 1.5 (unitless ratio)
Letter Spacing 0.5px, 0.02em "0.02em"

Naming Normalization

Source Style Target Style Example
color/primary/500 color.primary.500 Nested groups
Primary Color primaryColor camelCase
primary-color primaryColor camelCase
PRIMARY_COLOR primaryColor camelCase

Source Metadata

Preserve source information for traceability:

{
  "color": {
    "primary": {
      "$value": "#2196F3",
      "$type": "color",
      "$extensions": {
        "com.figma": {
          "variableId": "VariableID:1234",
          "styleId": "S:abc123",
          "collectionName": "Brand Colors"
        }
      }
    }
  }
}

Troubleshooting

No Tokens Found

Symptoms: Extraction returns empty or minimal tokens

Solution:

  1. Verify design file has published styles (not just local colors)
  2. Check if using Figma variables vs. older style system
  3. Look for design system component library
  4. Try extracting from specific frames/components

Authentication Failed

Symptoms: MCP server returns 401 or access denied

Solution:

  1. Verify FIGMA_ACCESS_TOKEN is set for Figma
  2. Ensure Sketch app is running for sketch-context
  3. Check Penpot MCP server is running locally

Inconsistent Naming

Symptoms: Token names don't match expected format

Solution:

  1. Check original design file naming conventions
  2. Use normalization rules consistently
  3. Consider mapping file to enforce naming standards

See Also

  • design-token-json style - W3C format output
  • design-token-swift style - Swift code output
  • design-to-swiftui skill - Generate SwiftUI views from tokens
  • design-system-management skill - Manage token libraries
Weekly Installs
1
Repository
arustydev/ai
GitHub Stars
6
First Seen
7 days ago
Installed on
zencoder1
amp1
cline1
openclaw1
opencode1
cursor1