nanobanana · maragudk — Claude Code skill: generate and edit images via CLI
A skill that shows your agent how to drive the nanobanana CLI: turning text prompts into images and editing files you already have.
MIT 1 installs ★ 49 updated 2026-03-27
npx skills add maragudk/skills --skill nanobanana -a claude-codeWhat nanobanana does
The skill is a written reference your coding agent reads before it touches the nanobanana CLI. That CLI talks to the Nano Banana image API, which runs on Google's generative models. Your job shrinks to a request; the agent picks the flags and builds the command.
Each call writes a file: nanobanana generate output.png "...", saved as .png or .jpg. Add -pro to reach the premium model — better output, a slower run. Editing an existing picture means passing -i with the source path and describing only the change you want.
Two things must exist first. Install the CLI with go install maragu.dev/nanobanana@latest, then provide GOOGLE_API_KEY, exported in your shell or written into a .env file in the directory you run from. Skip the key and the first call fails.
What it helps with
- generate a logo from a text brief
- recolour an existing photo with a prompt
- sketch concept art before a design discussion
- switch to the pro model when quality beats speed
- save PNG or JPG output straight from the terminal
Example prompts for the agent
- Generate
logo.pngas a minimalist mark with blue and white geometric shapes. - Edit
photo.png: purple sky, add stars, write the result tophoto-night.png. - Rerun that shot with
-pro, I want a cyberpunk skyline at night.
How to install nanobanana
1Skills CLI (Vercel)
npx skills add maragudk/skills --skill nanobanana -a claude-code2Manually: put the files in place
~/.claude/skills/nanobanana/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/nanobanana/
1Skills CLI (Vercel)
npx skills add maragudk/skills --skill nanobanana -a codex2Manually: put the files in place
.agents/skills/nanobanana/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add maragudk/skills --skill nanobanana -a cursor2Manually: put the files in place
.cursor/skills/nanobanana/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install maragudk/skills nanobanana2Manually: put the files in place
.github/skills/nanobanana/SKILL.md1Skills CLI (Vercel)
npx skills add maragudk/skills --skill nanobanana -a gemini-cli2Platform CLI
gemini skills install https://github.com/maragudk/skills --path nanobananaManually: put the files in place
.devin/skills/nanobanana/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add maragudk/skills --skill nanobanana -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about nanobanana
How do I install nanobanana in Claude Code?
Run npx skills add maragudk/skills --skill nanobanana -a claude-code. The skill lands at ~/.claude/skills/nanobanana/SKILL.md.
What does nanobanana do?
It gives the agent a guide to the nanobanana CLI: creating images from a description, modifying existing files, and when to use -pro.
What do I need before it works?
The CLI itself (go install maragu.dev/nanobanana@latest) plus GOOGLE_API_KEY as an environment variable or in .env inside the working directory.
Source text
SKILL.md · MIT
--- name: nanobanana description: Guide for generating and editing images using generative AI with the nanobanana CLI license: MIT --- # Nanobanana This is a guide for generating and editing images using the nanobanana CLI tool. Nanobanana is a command-line interface for the Nano Banana image generation API, which uses Google's generative AI models. ## Installation If the nanobanana CLI is not already installed, install it using: ```bash go install maragu.dev/nanobanana@latest ``` ## Prerequisites The `GOOGLE_API_KEY` environment variable must be set, or a `.env` file with the key must be present in the working directory. ## Generating images To generate a single image: ```bash nanobanana generate output.png "a beautiful sunset over mountains" ``` The output can be either a `.png` or `.jpg` file. ## Using the pro model To use the higher-quality pro model for generation: ```bash nanobanana generate -pro output.png "a beautiful sunset over mountains" ``` The `-pro` flag enables Google's premium generative AI model, which typically produces higher quality results but may take slightly longer to generate. ## Editing existing images To edit or modify an existing image using a text prompt: ```bash nanobanana generate -i input.png output.png "make the sky purple and add stars" ``` This is useful for making specific changes to existing images based on natural language instructions. ## Tips for effective prompts - Be specific and descriptive about what you want in the image - Include details about style, colors, mood, composition, and subject matter - For edits, clearly describe what should change while the rest remains the same - You can reference art styles, artists, or specific visual aesthetics ## Examples Generate a logo: ```bash nanobanana generate logo.png "minimalist tech company logo with geometric shapes in blue and white" ``` Create concept art: ```bash nanobanana generate concept.png "futuristic city skyline at night with neon lights and flying vehicles, cyberpunk style" ``` Edit an existing photo: ```bash nanobanana generate -i photo.png enhanced.png "enhance the colors and make it look like golden hour" ```
© maragudk · MIT · maragudk/skills
Similar skills
All skills for AI and ML →nanobanana by other authors comes first.