When to Use
User needs AI-generated images. Agent handles text-to-image, image editing, style transfer, upscaling, and provider selection.
Architecture
User preferences persist in ~/image-generation/. See memory-template.md for setup.
~/image-generation/
├── memory.md # Current provider, style, projects
└── history.md # Past generations (optional)
Quick Reference
| Topic |
File |
| Memory setup |
memory-template.md |
| Prompt techniques |
prompting.md |
| API handling |
api-patterns.md |
| OpenAI/DALL-E |
openai.md |
| Midjourney |
midjourney.md |
| Stable Diffusion |
stable-diffusion.md |
| Flux |
flux.md |
| Leonardo |
leonardo.md |
| Ideogram |
ideogram.md |
| Replicate |
replicate.md |
Core Rules
1. Check Memory First
Read ~/image-generation/memory.md for user's provider, preferred styles, and project context.
2. Draft Before Final
- Start at 512x512 or 1024x1024 to validate prompt
- Generate 4+ variations
- Only upscale the winner
3. Provider Selection by Task
| Task |
Best Provider |
| Photorealism |
Midjourney, Flux Pro |
| Text in images |
Ideogram, DALL-E 3 |
| Fast iteration |
Flux Schnell, Leonardo |
| Maximum control |
Stable Diffusion |
| Inpainting/editing |
DALL-E 3, Stable Diffusion |
| Budget API |
Replicate, Leonardo |
4. Prompt Structure
- Subject first: "A red fox" not "In the forest there is a red fox"
- Style keywords: "cinematic lighting", "oil painting", "studio photography"
- Be specific: "golden hour sunlight" not "good lighting"
- Match aspect ratio to content: 1:1 portraits, 16:9 landscapes
5. Update Memory
| Event |
Action |
| User chooses provider |
Save to memory.md |
| Style works well |
Note in memory.md |
| New project started |
Add to memory.md |
Common Traps
- Hands/fingers wrong → regenerate or use inpainting
- Text garbled → use Ideogram or add text in post-production
- Faces distorted → add "detailed face" to prompt, use face-fix models
- Style inconsistent → lock seed, use reference images
- Watermarks appearing → check model training, use clean models
Security & Privacy
Data that leaves your machine:
- Prompts sent to chosen AI provider for generation
Data that stays local:
- Provider preferences in
~/image-generation/
- No telemetry or analytics
This skill does NOT:
- Store generated images (provider handles storage)
- Access files outside
~/image-generation/
External Endpoints
| Provider |
Endpoint |
Data Sent |
Purpose |
| OpenAI |
api.openai.com |
Prompt text |
DALL-E generation |
| Midjourney |
discord.com |
Prompt text |
Image generation |
| Stability AI |
api.stability.ai |
Prompt text |
Stable Diffusion |
| Replicate |
api.replicate.com |
Prompt text |
Flux, SD models |
| Leonardo |
cloud.leonardo.ai |
Prompt text |
Leonardo generation |
| Ideogram |
api.ideogram.ai |
Prompt text |
Text-in-image |
Endpoints depend on chosen provider. No other data is sent externally.
Trust
By using this skill, prompts are sent to third-party AI providers (OpenAI, Midjourney, Stability AI, etc.).
Only install if you trust these services with your prompts.
Feedback
- If useful:
clawhub star image-generation
- Stay updated:
clawhub sync
1---2name: image-generation3description: Create AI images with prompt engineering, style control, and provider guides for Midjourney, DALL-E, Stable Diffusion, Flux, and Leonardo.4---56## When to Use78User needs AI-generated images. Agent handles text-to-image, image editing, style transfer, upscaling, and provider selection.910## Architecture1112User preferences persist in `~/image-generation/`. See `memory-template.md` for setup.1314```15~/image-generation/16├── memory.md # Current provider, style, projects17└── history.md # Past generations (optional)18```1920## Quick Reference2122| Topic | File |23|-------|------|24| Memory setup | `memory-template.md` |25| Prompt techniques | `prompting.md` |26| API handling | `api-patterns.md` |27| OpenAI/DALL-E | `openai.md` |28| Midjourney | `midjourney.md` |29| Stable Diffusion | `stable-diffusion.md` |30| Flux | `flux.md` |31| Leonardo | `leonardo.md` |32| Ideogram | `ideogram.md` |33| Replicate | `replicate.md` |3435## Core Rules3637### 1. Check Memory First38Read `~/image-generation/memory.md` for user's provider, preferred styles, and project context.3940### 2. Draft Before Final41- Start at 512x512 or 1024x1024 to validate prompt42- Generate 4+ variations43- Only upscale the winner4445### 3. Provider Selection by Task4647| Task | Best Provider |48|------|---------------|49| Photorealism | Midjourney, Flux Pro |50| Text in images | Ideogram, DALL-E 3 |51| Fast iteration | Flux Schnell, Leonardo |52| Maximum control | Stable Diffusion |53| Inpainting/editing | DALL-E 3, Stable Diffusion |54| Budget API | Replicate, Leonardo |5556### 4. Prompt Structure57- Subject first: "A red fox" not "In the forest there is a red fox"58- Style keywords: "cinematic lighting", "oil painting", "studio photography"59- Be specific: "golden hour sunlight" not "good lighting"60- Match aspect ratio to content: 1:1 portraits, 16:9 landscapes6162### 5. Update Memory63| Event | Action |64|-------|--------|65| User chooses provider | Save to memory.md |66| Style works well | Note in memory.md |67| New project started | Add to memory.md |6869## Common Traps7071- **Hands/fingers wrong** → regenerate or use inpainting72- **Text garbled** → use Ideogram or add text in post-production73- **Faces distorted** → add "detailed face" to prompt, use face-fix models74- **Style inconsistent** → lock seed, use reference images75- **Watermarks appearing** → check model training, use clean models7677## Security & Privacy7879**Data that leaves your machine:**80- Prompts sent to chosen AI provider for generation8182**Data that stays local:**83- Provider preferences in `~/image-generation/`84- No telemetry or analytics8586**This skill does NOT:**87- Store generated images (provider handles storage)88- Access files outside `~/image-generation/`8990## External Endpoints9192| Provider | Endpoint | Data Sent | Purpose |93|----------|----------|-----------|---------|94| OpenAI | api.openai.com | Prompt text | DALL-E generation |95| Midjourney | discord.com | Prompt text | Image generation |96| Stability AI | api.stability.ai | Prompt text | Stable Diffusion |97| Replicate | api.replicate.com | Prompt text | Flux, SD models |98| Leonardo | cloud.leonardo.ai | Prompt text | Leonardo generation |99| Ideogram | api.ideogram.ai | Prompt text | Text-in-image |100101Endpoints depend on chosen provider. No other data is sent externally.102103## Trust104105By using this skill, prompts are sent to third-party AI providers (OpenAI, Midjourney, Stability AI, etc.).106Only install if you trust these services with your prompts.107108## Feedback109110- If useful: `clawhub star image-generation`111- Stay updated: `clawhub sync`
johnalbertini14-glitch