web-design-studio（网页设计工作室）

This skill combines frontend design expertise with AI image generation to create complete, production-ready web pages with all visual elements included.

When to Use This Skill

Use this skill when the user asks for:

• Complete websites or landing pages
• Web dashboards or applications
• Product pages or portfolios
• Marketing pages with banners/hero sections
• Blog or content layouts with featured images
• Any web interface that needs visual assets

How This Skill Works

This skill performs a two-step process:

Step 1: Frontend Design

Follow the frontend-design aesthetic principles to create a distinctive, production-grade interface:

Design Thinking:

• Purpose: What problem does this interface solve? Who uses it?
• Tone: Pick a bold aesthetic direction (minimalist, maximalist, retro-futuristic, organic, luxury, playful, brutalist, etc.)
• Constraints: Technical requirements (framework, performance, accessibility)
• Differentiation: What makes this UNFORGETTABLE?

Frontend Aesthetics:

• Typography: Choose distinctive, beautiful fonts (avoid Inter, Roboto, Arial, system fonts)
• Color & Theme: Commit to a cohesive aesthetic with CSS variables
• Motion: Use animations for effects and micro-interactions
• Spatial Composition: Unexpected layouts, asymmetry, overlap, diagonal flow
• Backgrounds & Visual Details: Gradient meshes, noise textures, geometric patterns, dramatic shadows

Step 2: Image Generation

Identify all images needed in the design and generate them using the image generation script:

Note: The script paths below use ~/.claude/skills/ as the default installation location. If you installed this skill via npx skills add, the actual path may be different (e.g., ~/.local/share/npx/skills/ or similar). Adjust the path according to your setup.

Identify Image Needs:

• Hero/banner images
• Product/portfolio images
• Illustrations or icons
• Background images
• Section dividers or decorative elements

Generate Images: Run the image generation script for each image needed. Images are automatically saved to the project's images/ directory for better organization:

uv run ~/.claude/skills/web-design-studio/scripts/generate_image.py \
  --prompt "detailed image description matching the design aesthetic" \
  --filename "yyyy-mm-dd-hh-mm-ss-descriptive-name.png" \
  --resolution 1K|2K|4K

Resolution Guidelines:

• 1K (default) - ~1024px for small images, icons, thumbnails
• 2K - ~2048px for standard images, product shots
• 4K - ~4096px for hero banners, large backgrounds

Filename Format: {timestamp}-{descriptive-name}.png

• Timestamp: yyyy-mm-dd-hh-mm-ss (24-hour format)
• Name: lowercase descriptive text with hyphens

Image Organization:

• Images are automatically saved to the project's images/ directory
• The script detects git repository roots and places images at the project level
• When referencing images in code, use: images/filename.png

Step 3: Complete Implementation

After generating all needed images, create the final code with:

1. All generated images properly referenced using images/filename.png
2. Optimized image attributes (alt text, loading, sizes)
3. Responsive image handling
4. Complete styling matching the aesthetic vision

IMPORTANT: After creating the HTML file, you MUST verify that all images load correctly:

python ~/.claude/skills/web-design-studio/scripts/verify_images.py

This script checks:

• All image references in HTML files
• Whether the referenced image files exist
• Common path issues (missing images/ prefix, etc.)

Do not proceed to the next step until all images are verified successfully.

Step 4: Design Documentation (REQUIRED)

After image verification, you MUST generate a professional design specification document for client delivery. This is a mandatory final step.

Basic Usage:

python ~/.claude/skills/web-design-studio/scripts/generate_design_doc.py \
  --project-name "Coffee Shop Landing Page" \
  --output "design-doc-coffee-shop.html"

Auto-Extraction from HTML (RECOMMENDED):

python ~/.claude/skills/web-design-studio/scripts/generate_design_doc.py \
  --project-name "Coffee Shop Landing Page" \
  --output "design-doc-coffee-shop.html" \
  --html-file "index.html"

The --html-file parameter automatically extracts:

• Page sections with their actual content summaries
• Color scheme from CSS variables
• All images with filenames and resolutions
• CTA button count

This is the recommended approach as it saves time and ensures accuracy by analyzing the actual HTML code.

Advanced Usage (Manual Parameters):

python ~/.claude/skills/web-design-studio/scripts/generate_design_doc.py \
  --project-name "Coffee Shop Landing Page" \
  --output "design-doc-coffee-shop.html" \
  --design-concept "温暖有机的咖啡文化体验，采用大地色系和质朴风格" \
  --target-audience "咖啡爱好者、年轻职场人士、追求品质生活的消费者" \
  --primary-color "#8B4513" \
  --secondary-color "#A0522D" \
  --accent-color "#D4AF37" \
  --typography "标题：Playfair Display，正文：Lato" \
  --page-sections "Hero区,产品展示,品牌故事,客户评价,联系信息,页脚" \
  --cta-count "8" \
  --tech-stack "HTML5,CSS3,JavaScript,响应式设计" \
  --features "图片懒加载,平滑滚动,表单验证,移动优化" \
  --author "设计团队名称" \
  --project-type "landing-page" \
  --version "1.0"

With Image List:

python ~/.claude/skills/web-design-studio/scripts/generate_design_doc.py \
  --project-name "Coffee Shop" \
  --output "design-doc.html" \
  --image-list "2026-01-26-10-30-01-hero.png|4K|Banner区背景,2026-01-26-10-30-02-product.png|2K|产品展示" \
  --primary-color "#0a2540" \
  --secondary-color "#1e3a5f" \
  --accent-color "#ff6b2c"

Document Contents (6 Major Sections):

• 01 设计概述 - 核心理念、设计目标、目标受众
• 02 视觉设计策略 - 色彩系统、字体设计、图片策略
• 03 页面结构设计 - 完整页面布局和区块说明
• 04 用户体验设计 - 交互设计、响应式设计、性能优化
• 05 技术实现 - 技术栈、代码特点、浏览器兼容
• 06 交付清单 - 已交付内容、图片清单、建议后续工作

Output: A professionally formatted HTML document with sticky navigation, comprehensive sections, and beautiful styling - perfect for client presentations and design handoffs.

Key Parameters:

• --html-file: (RECOMMENDED) Auto-extract sections, colors, images, CTAs from HTML
• --design-concept, --target-audience: Project description
• --primary-color, --secondary-color, --accent-color, --bg-color: Custom color scheme
• --page-sections: Comma-separated section list
• --image-list: Format: filename1|resolution1|purpose1,filename2|resolution2|purpose2
• --cta-count: Number of CTA buttons
• --author: Design team/author name
• --project-type: landing-page, website, dashboard, etc.
• --version: Document version number

Complete Workflow Example

User Request: "Create a landing page for a coffee shop"

Process:

1. Design Phase: Choose a warm, organic aesthetic with earth tones, rustic typography
2. Image Planning: Identify needs for:

- Hero banner with coffee atmosphere - Product shots for coffee cups - Background textures for sections

1. Generate Images: # Hero banner (auto-saved to images/) uv run ~/.claude/skills/web-design-studio/scripts/generate_image.py \ --prompt "warm rustic coffee shop interior, golden hour lighting" \ --filename "2026-01-24-10-30-15-coffee-shop-hero.png" \ --resolution 4K # Product shot (auto-saved to images/) uv run ~/.claude/skills/web-design-studio/scripts/generate_image.py \ --prompt "artisan coffee cup on wooden table, latte art, top view" \ --filename "2026-01-24-10-32-45-coffee-cup-product.png" \ --resolution 2K # Background texture (auto-saved to images/) uv run ~/.claude/skills/web-design-studio/scripts/generate_image.py \ --prompt "subtle coffee bean pattern texture, warm brown gradient" \ --filename "2026-01-24-10-35-20-coffee-texture-bg.png" \ --resolution 2K
2. Create Code: Build complete HTML/CSS with all images integrated using images/filename.png
3. Verify Images: (CRITICAL STEP) python ~/.claude/skills/web-design-studio/scripts/verify_images.py This ensures all image paths are correct and files exist. Fix any issues before proceeding.
4. Generate Design Doc: (REQUIRED FINAL STEP) python ~/.claude/skills/web-design-studio/scripts/generate_design_doc.py \ --project-name "Coffee Shop Landing Page" \ --output "design-doc-coffee-shop.html" \ --html-file "index.html" The script will automatically extract: Optional: Add custom parameters for more control: python ~/.claude/skills/web-design-studio/scripts/generate_design_doc.py \ --project-name "Coffee Shop Landing Page" \ --output "design-doc-coffee-shop.html" \ --html-file "index.html" \ --design-concept "温暖有机的咖啡文化体验，采用大地色系和质朴风格" \ --target-audience "咖啡爱好者、年轻职场人士" \ --author "设计团队"

- Page sections with content summaries - Color scheme from CSS - All images and their resolutions - CTA button count

API Key

The script requires a Gemini API key:

• Pass via --api-key argument if user provides one
• Otherwise uses GEMINI_API_KEY environment variable

Best Practices

1. Design First, Images Second: Plan the complete design before generating images
2. Match Aesthetics: Ensure generated images match the chosen design aesthetic
3. Optimize Resolution: Use appropriate resolution for each image's purpose
4. Descriptive Prompts: Write detailed, aesthetic-specific prompts for better results
5. Organize Filenames: Use clear, descriptive filenames matching the design context
6. Verify Images: ALWAYS run the verification script after creating HTML to catch path issues early
7. Generate Documentation: Design documentation is MANDATORY - never skip this final step

Output

This skill produces the following complete deliverables:

1. Complete HTML/CSS/JS Code: Production-ready, responsive, fully functional
2. Generated Images: All images saved to images/ directory with proper naming
3. Verified Image References: All image paths validated and working (via verify_images.py)
4. Design Documentation: MANDATORY professional spec document (design-doc-{project}.html)
5. Clean Project Structure: Organized files ready for deployment

Verification Checklist Before Completion:

• All images generated and in images/ directory
• HTML file created with image references as images/filename.png
• Verification script passes: python verify_images.py
• Design documentation generated: python generate_design_doc.py
• All deliverables are present and accessible

Remember: The goal is to create COMPLETE web experiences where every visual element is thoughtfully designed and generated - no placeholder images, no "image goes here" sections. Real, finished work.