| name | ai-creative-strategist |
| description | Research-powered creative strategy with visual previews. Use when you need to figure out WHAT to create before creating it. Researches competitors, captures references, generates visual direction previews, and outputs comprehensive briefs. Triggers on: creative strategy, what visuals should I create, creative brief, visual direction. Outputs briefs with actual preview images for feedback. |
AI Creative Strategist
The difference between amateur and professional creative isn't executionโit's strategy.
This skill researches, strategizes, and generates visual previews before full production. You see the direction before committing.
MCP Setup (First Time Only)
These skills use MCP servers for image generation and visual research. If you don't have them installed yet, here's how to set them up.
Check What's Installed
Ask Claude: "What MCP tools do you have available?"
If you see glif or replicate tools โ you're ready for generation.
If you see playwright or browser_ tools โ you're ready for visual research.
Image Generation (Pick One or Both)
Option A: Glif (recommended for beginners)
- Get your API token at glif.app โ Settings โ API
- Add to your Claude Desktop config (
claude_desktop_config.json):
"glif": {
"command": "npx",
"args": ["-y", "@glifxyz/glif-mcp-server@latest"],
"env": {
"GLIF_API_TOKEN": "your-token-here"
}
}
Option B: Replicate (more models, pay-per-use)
- Get your API token at replicate.com โ Account โ API tokens
- Add to your Claude Desktop config:
"replicate": {
"command": "npx",
"args": ["-y", "replicate-mcp"],
"env": {
"REPLICATE_API_TOKEN": "your-token-here"
}
}
Visual Research (Optional but Recommended)
Playwright โ captures competitor screenshots for reference
"playwright": {
"command": "npx",
"args": ["-y", "@playwright/mcp@latest"]
}
Where's the Config File?
Claude Desktop (Mac):
~/Library/Application Support/Claude/claude_desktop_config.json
Claude Desktop (Windows):
%APPDATA%\Claude\claude_desktop_config.json
After editing, restart Claude Desktop to load the new MCPs.
Claude Code Users
Add to your ~/.claude/settings.json under mcpServers, same format as above.
The Process
1. DISCOVER โ Diagnose creative needs
2. RESEARCH โ Gather competitive intelligence
3. PREVIEW โ Generate visual direction samples โ NEW
4. FEEDBACK โ Refine based on what you see
5. BRIEF โ Output comprehensive creative brief
When to Use This Skill
Use when:
- Starting a new visual campaign or brand
- Launching a product and need visual direction
- Current visuals aren't performing
- Need to differentiate from competitors
- Don't know what style/aesthetic to pursue
Skip to execution when:
- You have a clear brief already
- Simple, one-off generations
Phase 1: Discovery
Quick diagnosis before research:
## Creative Strategy Intake
**What are we creating for?**
[ ] Product launch
[ ] Brand refresh
[ ] Campaign (ads, social)
[ ] Website/landing page
[ ] Social media presence
**Who is the audience?**
- Primary: _______
- What do they respond to visually?
**Competitors:**
- Direct: _______
- Aspirational brands: _______
**Budget/Volume:**
[ ] Low ($10-50)
[ ] Medium ($50-200)
[ ] High ($200+)
Phase 2: Research (MCP-Powered)
Competitor Visual Analysis (Playwright)
Capture live screenshots as references:
1. browser_navigate to competitor site
2. browser_snapshot to find elements
3. browser_take_screenshot for hero, products
4. Analyze: colors, style, lighting, mood
Trend Research (Perplexity)
perplexity_research:
"What are the visual trends for [industry] in 2025?
- Color palettes
- Photography styles
- Overall aesthetic direction"
Competitive Analysis (Firecrawl)
firecrawl_scrape competitor sites:
- Homepage messaging
- Product positioning
- Brand voice
Phase 3: Visual Direction Previews โญ NEW
Before finalizing the brief, generate actual images to see the direction.
Generate 2-3 Direction Options
Based on research, create distinct visual directions:
Direction A: [Name - e.g., "Premium Minimal"]
{
"model_owner": "google",
"model_name": "nano-banana-pro",
"input": {
"prompt": "[Product/concept] with [style A characteristics], [lighting], [mood], professional commercial quality",
"aspect_ratio": "16:9"
}
}
Direction B: [Name - e.g., "Bold & Energetic"]
{
"model_owner": "google",
"model_name": "nano-banana-pro",
"input": {
"prompt": "[Product/concept] with [style B characteristics], [lighting], [mood], professional commercial quality",
"aspect_ratio": "16:9"
}
}
Direction C: [Name - e.g., "Warm Lifestyle"]
{
"model_owner": "google",
"model_name": "nano-banana-pro",
"input": {
"prompt": "[Product/concept] with [style C characteristics], [lighting], [mood], professional commercial quality",
"aspect_ratio": "16:9"
}
}
Present Visual Options
## Visual Direction Previews
Based on research, here are 3 distinct directions:
### Direction A: Premium Minimal
- **Characteristics:** Clean, dark backgrounds, dramatic lighting
- **Mood:** Sophisticated, premium, confident
- **Differentiation:** Stands out from [competitor] with [contrast]
### Direction B: Bold & Energetic
- **Characteristics:** Vibrant colors, dynamic angles, high contrast
- **Mood:** Exciting, modern, disruptive
- **Differentiation:** Opposite of category convention of [typical approach]
### Direction C: Warm Lifestyle
- **Characteristics:** Natural lighting, contextual settings, warm tones
- **Mood:** Approachable, authentic, trustworthy
- **Differentiation:** Human touch vs [competitor's] cold tech feel
---
**Which direction resonates?**
- Pick one to pursue
- Combine elements from multiple
- Request variations on a direction
Phase 4: Feedback Loop
Based on visual previews, refine:
User: "I like Direction A's lighting but Direction C's warmth"
โ Generate refined preview combining elements:
{
"prompt": "[Product] with dramatic lighting from Direction A,
warm color grading from Direction C, [specific adjustments]..."
}
โ Present refined preview
โ Continue iterating until direction is locked
This is the key differentiator: See actual images, give feedback, refine before committing to full production.
Phase 5: Final Brief
Once direction is approved via previews:
# Creative Brief: [Project Name]
## Approved Direction
**Selected:** [Direction name]
**Preview that was approved:** [URL]
## Visual Direction (Validated)
### Mood & Tone
[Based on approved preview]
### Color Direction
- Primary: [extracted from approved preview]
- Secondary: [accents visible in preview]
### Style Characteristics
[What made the approved preview work]
### Prompt Template (Proven)
[The exact prompt that generated the approved direction,
ready to scale to other assets]
## Asset Requirements
| Asset | Ratio | Notes |
|-------|-------|-------|
| Hero image | 16:9 | [based on preview] |
| Product shots | 1:1 | [same style] |
| Social graphics | 4:5 | [adapted] |
## Model Recommendation
**Nano Banana Pro** for all production (same as previews)
## Budget
- Previews generated: $0.06 (3 ร $0.02)
- Production remaining: ~$X for [X assets]
---
**Direction validated. Ready for production.**
โ Route to execution skills
Visual Preview Patterns
Pattern 1: Product Launch
Generate previews for:
1. Hero shot (product floating, dramatic)
2. Lifestyle shot (product in context)
3. Abstract/mood shot (brand feeling)
Let user pick which approach for full campaign.
Pattern 2: Brand Refresh
Generate previews showing:
1. Evolution of current style
2. Complete departure
3. Competitive differentiation angle
Validate new direction with stakeholders.
Pattern 3: Campaign Creative
Generate previews for:
1. Ad creative concept A
2. Ad creative concept B
3. Ad creative concept C
Test different angles before production spend.
Research โ Preview โ Brief Flow
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ 1. RESEARCH โ
โ - Perplexity: Trends โ
โ - Playwright: Competitor screenshots โ
โ - Firecrawl: Messaging analysis โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ
โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ 2. VISUAL PREVIEWS (Nano Banana Pro) โ
โ - Generate 2-3 direction options โ
โ - Present to user with analysis โ
โ - Gather feedback on what works โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ
โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ 3. ITERATE โ
โ - Refine based on feedback โ
โ - Generate new previews if needed โ
โ - Lock direction when approved โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ
โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ 4. FINAL BRIEF โ
โ - Approved preview as reference โ
โ - Proven prompt template โ
โ - Ready for scaled production โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ
โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ 5. EXECUTION (other skills) โ
โ - ai-product-photo โ
โ - ai-social-graphics โ
โ - ai-product-video โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Model for Previews
| Model | Use | Speed | Cost |
|---|
| Nano Banana Pro | All previews | ~3min | $0.02 |
Why Nano Banana Pro for previews:
- Same quality as production (what you see is what you get)
- Consistent results when scaling approved direction
- Worth the 3min wait for strategic decisions
Example: Full Flow
User: "I need visuals for a new AI productivity tool targeting developers"
Step 1: Research
Competitors analyzed: GitHub Copilot, Cursor, Linear
Trend research: Dark mode, minimal, purple/blue palettes
Whitespace found: "Warm + Bold" quadrant underserved
Step 2: Generate Previews
Direction A: "Dark Sophisticated"
Premium AI interface visualization against dark charcoal gradient,
subtle purple accent lighting, clean minimal composition,
tech startup aesthetic, professional commercial quality
Direction B: "Warm Approachable"
AI productivity concept with warm coral accents on dark background,
soft gradient lighting, friendly but professional feel,
modern developer tool aesthetic, commercial quality
Direction C: "Bold Disruptive"
AI concept with vibrant gradient colors, dynamic composition,
high energy modern design, standing out from typical dev tools,
bold startup aesthetic, commercial quality
Step 3: Present Previews
## Visual Direction Previews
### Direction A: Dark Sophisticated
[Image URL]
Aligns with: Linear, Vercel
Differentiation: Premium positioning
### Direction B: Warm Approachable โญ
[Image URL]
Aligns with: Raycast warmth + Linear sophistication
Differentiation: "Warm" in a cold category
### Direction C: Bold Disruptive
[Image URL]
Differentiation: Maximum contrast to competitors
Risk: May feel less "developer credible"
Step 4: User Feedback
User: "I love Direction B - that warmth feels right. Can we try it with more contrast?"
Generate refined preview:
AI productivity concept with warm coral accents (#FF6B6B) on deep charcoal,
higher contrast dramatic lighting, warm but powerful feel,
sophisticated developer tool aesthetic, commercial quality
Step 5: Lock & Brief
# Creative Brief: DevAssist AI
## Approved Direction
**Selected:** Warm Approachable (refined)
**Approved Preview:** [URL]
## Visual Direction
- Warm coral (#FF6B6B) on dark charcoal (#1A1A2E)
- High contrast but warm mood
- "Intelligent warmth" positioning
## Proven Prompt Template
[Product/concept] with warm coral accents on deep charcoal,
high contrast dramatic lighting, warm but powerful feel,
sophisticated developer tool aesthetic, commercial quality
## Ready for Production
โ Route to ai-product-photo, ai-social-graphics
Quality Gate
Before finalizing:
Handoff
## Creative Strategist Handoff
**Approved Direction Preview:** [URL]
**Proven Prompt Template:** [prompt that worked]
**Research Summary:**
- Competitive landscape: [key insights]
- Differentiation: [why this direction wins]
**Model:** Nano Banana Pro (same as previews)
**Route to:** ai-product-photo / ai-social-graphics / ai-product-video
