🍌
nano-banana-prompting
Structured prompts for Gemini image gen
v1.0.0
MIT License
✓ OpenClaw
➜~clawhub install nano-banana-prompting
What it does
JSON-structured prompting system for Gemini's native image generation. Produces consistent, high-quality results every time with repeatable prompt patterns.
How it works
Transforms your natural language into optimized structured JSON prompts for Gemini image generation. The difference is dramatic: a plain prompt like "a cat in space" produces generic AI slop, while the structured version produces a cinematic masterpiece with controlled lighting, composition, and style. The skill auto-detects the best style (cinematic, watercolor, pixel art, etc.) and builds the full JSON prompt automatically. No scripts, no dependencies — it teaches the agent HOW to prompt.
Features
🎨 Auto-detects style: cinematic, watercolor, pixel art, oil painting, and more
📐 Structured JSON prompts with subject, environment, lighting, mood, camera
🔄 Before/After: plain prompt → AI slop, structured prompt → masterpiece
🧱 Zero scripts, zero dependencies — pure knowledge injection
✨ Works with Gemini 2.0 Flash and Pro image generation
🎯 Repeatable results — same prompt structure = consistent quality
Example
output
❌ Plain prompt:
"a gecko on a computer"
→ Generic, flat, inconsistent lighting
✅ Structured prompt:
{
"subject": "small gecko perched on mechanical keyboard",
"environment": "dark room, multiple monitors glowing",
"lighting": "cool blue monitor glow, warm desk lamp",
"mood": "cozy, late-night coding session",
"style": "cinematic photography",
"camera": "macro lens, f/2.8, shallow depth of field",
"details": "RGB keyboard lights reflecting on gecko scales"
}
→ Cinematic, moody, publication-quality result