Gemini Imagegen

A solid implementation of Gemini's image generation API with the kind of refinement workflow you actually want when iterating on visuals. Supports up to 14 reference images for composition work, configurable resolutions from 1K to 4K, and ten different aspect ratios. The multi-turn chat interface means you can keep tweaking without starting over, which is how this stuff should work. Google Search grounding is interesting for generating images that need to reference current events or real-world subjects. Default model is gemini-3-pro-image-preview and outputs JPEG. You'll need a GEMINI_API_KEY environment variable. Good option if you're already in the Google ecosystem or want an alternative to the usual Stable Diffusion or DALL-E routes.

Install to Claude Code

npx -y skills add everyinc/compound-engineering-plugin --skill gemini-imagegen --agent claude-code