---

name: generate-image description: Generates images using Google Gemini 3 Pro API. Use when asked to create, generate, or design images for slides, presentations, or visual content. allowed-tools: Bash(bun:*), Read, Write

Image Generation with Gemini 3 Pro

Generate images for slides and presentations using Google's Gemini 3 Pro image generation API.

Prerequisites

Ensure .env.local exists with your Gemini API key:

CopyGEMINI_API_KEY=your-api-key-here

Usage

Run the generation script:

Copybun --env-file=.env.local .claude/skills/generate-image/scripts/generate.ts --prompt "your image description" --ratio "16:9" --output "output.png"

Parameters

Examples

Copy# Widescreen slide background
bun --env-file=.env.local .claude/skills/generate-image/scripts/generate.ts \
  --prompt "Abstract dark terminal with glowing code, cyberpunk aesthetic" \
  --ratio "16:9"

# Square icon
bun --env-file=.env.local .claude/skills/generate-image/scripts/generate.ts \
  --prompt "Minimalist code icon, orange and green on dark background" \
  --ratio "1:1" \
  --output "code-icon.png"

Output

Images are saved to public/ with auto-generated timestamps if no filename specified.

Prompts are automatically logged to src/prompts/ as JSON files containing:

• filename: Output image filename
• prompt: The prompt used to generate the image
• ratio: Aspect ratio used
• timestamp: Generation timestamp (ISO format)
• model: Gemini model used

This allows you to reference and regenerate images later using the same prompts.

Troubleshooting

• API key not found: Ensure .env.local exists with GEMINI_API_KEY
• Generation failed: Check API key validity and quota
• Model error: The script uses gemini-3-pro-image-preview model