nano-banana-2 (OpenRouter edition)

Overview

This skill generates / edits images via OpenRouter (OpenAI-compatible API) using:

• model: google/gemini-3.1-flash-image-preview

Compared to a raw curl example, this version is more reliable because it:

• sends multimodal messages (text + optional input images)
• parses the returned base64 data:image/... payloads
• writes output images to disk
• prints MEDIA: <path> lines (so other tools/agents can pick them up)

One-time setup

Ensure OPENROUTER_API_KEY is set in the environment.

Install Python dependency (if missing):