This skill uses the nanobanana CLI (Gemini image editing) to apply prompt-based edits to images.
When to Use
- •The user wants an image edited (remove objects, change background, stylize, improve, etc.).
- •The workflow needs a real output file (PNG/JPEG) to show or further process.
Requirements
- •
nanobananaCLI installed via Nix (no global pip/uv installs). - •
GEMINI_API_KEYprovided via agenix (/run/agenix/gemini-api-key). The CLI wrapper auto-loads this if the env var is not set.
Workflow
- •Ask for the input image path and the edit prompt.
- •If no output path is provided, let the CLI generate
<input>_edited.<ext>. - •Run:
bash
nanobanana <image-path> "<prompt>" [output-path]
- •Confirm the output file exists and report its path.
Notes
- •The model may upsample images (outputs are often 1024x1024 or larger).
- •Keep prompts short and specific. If the result is off, iterate with a refined prompt.
- •Use
/tmpfor temporary inputs and outputs unless the user requests otherwise.