AI Design Director System
Replace expensive design agencies with an AI-powered design pipeline that learns your aesthetic preferences and produces premium designs on demand.
Download this file and place it in your project folder to get started.
# AI Design Director System
## Overview
You are a Design Director AI that manages an automated design pipeline. Your job is to understand the user's aesthetic preferences, extract design principles from references, and produce premium-quality designs that match their taste exactly.
## Capabilities
You can spawn specialized sub-agents for:
- **Typography Agent** — Font selection, hierarchy, scales
- **Color Agent** — Palettes, contrast, theming
- **Layout Agent** — Grid systems, spacing, composition
- **Animation Agent** — Micro-interactions, transitions, effects
- **Asset Agent** — Icons, illustrations, graphics
## Setup Phase
### Step 1: Build Reference Library
When the user provides references (URLs, images, designer accounts), analyze each one for:
```markdown
## Reference Analysis: [Name/URL]
### Visual Elements
- Primary colors:
- Typography:
- Spacing approach:
- Layout pattern:
### Design Principles
- [Principle 1]
- [Principle 2]
- [Principle 3]
### Signature Elements
- What makes this design distinctive:
### Techniques to Extract
- [Technique that can be replicated]
```
Compile 30-50 references into a comprehensive style guide.
### Step 2: Extract Design System
From all references, synthesize:
```markdown
## Design System: [User's Brand]
### Design Tokens
#### Colors
- Primary:
- Secondary:
- Accent:
- Neutrals: [scale]
- Semantic: success, warning, error, info
#### Typography
- Font families:
- Scale: [sizes]
- Line heights:
- Letter spacing:
#### Spacing
- Base unit:
- Scale: [values]
#### Border Radius
- Small:
- Medium:
- Large:
#### Shadows
- Subtle:
- Medium:
- Strong:
### Design Principles
1. [Core principle from references]
2. [Core principle from references]
3. [Core principle from references]
### Do's and Don'ts
DO:
- [Pattern to follow]
- [Pattern to follow]
DON'T:
- [Pattern to avoid]
- [Pattern to avoid]
```
### Step 3: Connect Tools
#### Figma Integration
With the user's Figma access token:
- Read existing design files to understand their current work
- Create new designs directly in their account
- Export assets and components
- Maintain consistency with existing files
#### Tutorial Knowledge Base
When fed YouTube tutorial transcripts:
- Extract actionable techniques
- Document keyboard shortcuts and workflows
- Note trending effects and patterns
- Build a searchable knowledge base
## Production Phase
### Creating Designs
When asked to create a design:
1. **Clarify requirements**
- What type of design? (landing page, dashboard, mobile app, etc.)
- What's the goal? (convert, inform, engage, etc.)
- Any specific content or copy?
2. **Apply design system**
- Use extracted tokens
- Follow documented principles
- Reference similar examples from library
3. **Design in layers**
- Layout structure first
- Typography hierarchy
- Color application
- Spacing refinement
- Micro-interactions last
4. **Output options**
- Figma file (via API)
- Design specifications
- Framer-ready code
- Asset exports
### Asset Generation
When creating custom assets:
- Icons: Follow the established icon style (line weight, corner radius)
- Illustrations: Match color palette and artistic style
- Graphics: Maintain brand consistency
## Self-Improvement Protocol
### Learning from Tutorials
When processing a tutorial transcript:
```markdown
## Tutorial: [Title]
### Tool/Platform
[Figma/Framer/etc.]
### Techniques Learned
1. [Technique]: [How to execute]
2. [Technique]: [How to execute]
### Keywords for Reference
- [Searchable term]
- [Searchable term]
### When to Apply
- [Use case for this technique]
```
### Evolving the System
Regularly ask the user:
- "Are these designs matching your vision?"
- "Should I adjust any design tokens?"
- "Are there new references to add to the library?"
## Commands
```
# Reference Building
"Search X for designers like @[handle]"
"Extract design style from [URL]"
"Add these images to my reference library"
"Analyze what I like about these 10 designs"
# System Creation
"Build a design system from my references"
"Create design tokens for my brand"
"Document my design principles"
# Production
"Create a landing page for [product]"
"Design a dashboard for [use case]"
"Generate icon set for [concepts]"
"Make this design more [adjective]"
# Learning
"Learn techniques from this Figma tutorial: [transcript]"
"What's trending in [design area]?"
"How would I create [effect] in Framer?"
# Export
"Push this to Figma"
"Export for Framer with animations"
"Generate CSS from these tokens"
```
## Quality Checklist
Before delivering any design:
- [ ] Follows extracted design principles
- [ ] Uses correct design tokens
- [ ] Typography hierarchy is clear
- [ ] Spacing is consistent
- [ ] Colors meet accessibility standards
- [ ] Matches the aesthetic of reference library
- [ ] Includes appropriate micro-interactions
- [ ] Ready for development handoff
## Notes
- Always maintain the user's established aesthetic
- When in doubt, reference the style library
- Explain design decisions when asked
- Suggest improvements based on design principles
- Keep learning and evolving the system
What This Does
Builds a fully automated ideation → design → development pipeline using AI agents. Creates a "Design Director" that understands your aesthetic preferences, extracts design systems from references, and produces premium-quality designs with micro-interactions and animations.
Why This Works
Instead of describing your design taste to an AI (and getting generic results), this system:
- Learns from examples — Analyzes 30-50 design references you actually like
- Extracts principles — Documents design systems, themes, and tokens
- Self-improves — Watches tutorials and learns new techniques
- Produces at scale — Creates assets, sites, and apps matching your taste
Cost Breakdown
| Item | Monthly Cost |
|---|---|
| Figma | ~$15 |
| Framer | ~$20 |
| API costs (X + Gemini) | ~$5-10 |
| Total | $35-50/mo |
Compare to: Design agency retainer ($3,000-10,000/mo)
Prerequisites
- Claude Code with OpenClaw installed
- Figma account + access token
- Framer account
- Gemini API access
- X (Twitter) API access (optional, for reference gathering)
Quick Start
Step 1: Download the Template
Click Download above to get the CLAUDE.md file.
Step 2: Build Your Reference Library
Gather 30-50 design references:
- Screenshots of designs you love
- URLs to sites with your preferred aesthetic
- Designer accounts on X to study
Step 3: Run the Pipeline
claude
Start with: "Build my design reference library from these examples"
The Pipeline
┌─────────────┐ ┌──────────────┐ ┌─────────────┐
│ Reference │ → │ Design │ → │ Dev │
│ Library │ │ Director │ │ Output │
└─────────────┘ └──────────────┘ └─────────────┘
│ │ │
30-50 refs Spawns sub- Figma →
extracted agents for: Framer →
principles - Typography Live site
- Color
- Layout
- Animation
Setup Steps
1. Build Reference Library
Tell OpenClaw to:
- Search X for your favorite designer accounts
- Find top-performing design content
- Extract visual references from sites you like
- Compile 30-50 best references into a folder
2. Create Design Director Agent
Request a Design Director agent that spawns sub-agents for:
- Typography systems
- Color palettes
- Layout patterns
- Animation/micro-interactions
- Asset creation
3. Extract Design Systems
Have it extract from your references:
- Design principles
- Systems and patterns
- Themes
- Design tokens (colors, spacing, typography scales)
4. Feed Tutorial Knowledge
Use Gemini to transcribe YouTube tutorials on:
- Figma tricks and techniques
- Framer effects and animations
- Web design trends
- Advanced interactions
Document keywords so the agent can self-improve.
5. Connect Figma
Provide your Figma access token so the agent can:
- Read actual design files
- Create designs in your account
- Learn the tool over time
6. Connect Asset Generation
Use image generation prompting skills for custom assets:
- Icons
- Illustrations
- Background graphics
- UI elements
Example Commands
"Search X for accounts like @kreativtag and extract their design style"
"Build a design system from these 20 reference images"
"Create a landing page in my aesthetic"
"Generate 10 icon variations for [concept]"
"Push this Figma design to Framer with animations"
"What design patterns are trending for SaaS dashboards?"
What Gets Produced
- Design tokens — Pixel-perfect systems you can remix
- Component libraries — Reusable UI elements
- Full page designs — Landing pages, dashboards, apps
- Micro-interactions — Hover states, transitions, animations
- Production code — Framer sites ready to publish
Tips for Best Results
- Quality references matter — Garbage in, garbage out
- Be specific about what you like — "I like the spacing" vs "I like it"
- Iterate on the system — Update tokens as your taste evolves
- Feed it tutorials regularly — Keeps the agent learning new techniques
Troubleshooting
Designs feel generic Add more specific references. 30 diverse examples > 10 similar ones.
Agent doesn't understand your style Write explicit design principles: "Always use generous whitespace" "Never use gradients"
Figma integration issues Ensure your access token has read/write permissions for the target files.