Banana Claude skill
A Claude Code skill for generating, editing and reviewing images with Google Gemini models, with a plan and approval before every paid request
- 1,213 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Makes paid requests to the external Gemini API, and each attempt costs money
- Handles a sensitive Google key and writes generated files to disk
Install
- Works in
Manual install
/plugin marketplace add AgriciDaniel/banana-claude
/plugin install banana-claude@banana-claude-marketplace
/plugin enable banana-claude@banana-claude-marketplace
/reload-pluginsThe plugin installs disabled; enable it and set the Gemini key in the plugin settings.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Makes paid requests to the external Gemini API, and each attempt costs money
- Handles a sensitive Google key and writes generated files to disk
This is third-party code. Review the repository files before installing.
What it does
The skill turns a plain-language request into a visual brief, compiles the exact prompt, shows the plan and an estimate, and only after approval sends a paid request to Google Gemini models. It can generate images, make one change while protecting geometry and details, continue a session with prior settings, compare up to three prompts across model routes, and review a finished result for composition, text and rights. The Gemini key never reaches commands, URLs or metadata, and one approval permits one provider attempt. A result is not treated as done until the user inspects the actual pixels. A metadata sidecar is saved next to the image without the raw prompt.
Who it is for. For designers, marketers and content creators who generate and edit images directly in Claude Code.
Good fit when
- You need to generate a visual from a text description
- You need to change one detail in an image cleanly
- You need to compare several prompts and models in one pass
- You need to review a finished result for composition, text and rights
Not a fit when
- You have no Gemini key and no billing on a Google AI project
- You need free image generation
- You have no access to Google services from your network
Example request
Generate a 16:9 GitHub hero with clean copy space on the leftLimitations
You need a Gemini key for a billing-enabled Google AI project, and every request is paid. It requires Claude Code 2.1.199 or newer and Python 3.11 with python3 on PATH. Google AI services are unavailable from Russia without a VPN, so the skill will not work without one. Gemini output is probabilistic: the skill does not guarantee text accuracy, geometry, clean rights or the final invoice, and those stay on the user's review.
How to disable. Disable or remove the banana-claude plugin via /plugin. For a manual install, delete the banana skill folder from the skills directory and the banana MCP entry from the configuration.
FAQ
How much does it cost?
Every Gemini request is paid at Google's rates. The skill shows the plan and estimate before payment, while planning and prompt work are free.
Where is the key stored?
The key is set as a sensitive plugin value and is never written into commands, URLs or metadata.
How is Banana Claude different from generating images in Gemini directly?
The skill turns a plain request into a visual brief, builds a precise prompt, shows the plan and a cost estimate and only after your confirmation makes the paid request to Google Gemini models. It can generate, edit and review images.
How do I install Banana Claude and how much does generation cost?
In Claude Code: /plugin marketplace add AgriciDaniel/banana-claude, then /plugin install. Paid requests go through your Gemini key, and the skill shows the price before the request. Risk is high because of paid calls.
Discussion
No questions or reviews yet.
Related
SkillsThe media-pipeline plugin generates images through Gemini or OpenAI gpt-image-2: a skill, CLI and MCP server install in one step.
Nano Banana (cc-nano-banana)
Nano Banana
Claude Code draws and edits images, icons, diagrams and patterns through the nanobanana extension for Gemini CLI
A CLI and Claude Code skill for Gemini images from 512 to 4K, with reference-based edits and transparent assets
Skills give Claude Code three ways to make images: code-drawn graphics, Three.js 3D scenes and Flux diffusion on Cloudflare.
A generative art skill for Claude Code: deterministic hash-seeded sketches, traits and rarity, export for screens, print and plotters
The skill turns paper text, methods and captions into structured prompts for scientific figures, tuned to the chosen image model.
README in short
The README describes a skill that runs Gemini image generation through explicit steps: ask, plan, review the plan, approve, create, inspect the pixels. The plugin installs via /plugin marketplace add and comes disabled because generation is paid. The Gemini key is set as a sensitive value and never reaches commands or logs. It routes across the Nano Banana 2 Lite, Nano Banana 2 and Nano Banana Pro models. There is a separate guide, prompt and model references, and compare and typeset modes. MIT licensed, version 3.
README badge
Original SKILL.md
--- name: banana description: "Direct, generate, edit, compare, and review visual assets with current Google Gemini image models. Use for image creation, image editing, reference-based consistency, product and character visuals, text-bearing graphics, grounded diagrams, video-derived images, and multi-model image portfolios." argument-hint: "[generate|edit|continue|portfolio|typeset|preset|cost|doctor] <request>" metadata: version: "3.0.0" author: AgriciDaniel provider: Google Gemini Developer API --- # Banana Claude Turn user intent into a frozen visual brief, compile the exact prompt, plan the request, obtain approval, execute through the bundled Gemini client, and inspect the actual pixels. The prompt is a control artifact, not the finished work. Plugin command: /banana-claude:banana. The standalone install uses /banana and the direct scripts, without plugin MCP or plugin-managed secrets. ## Non-negotiable boundaries - Planning, prompt work, model inspection, and cost estimation do not call Google. Planning does write a short-lived approval capability to private local state. - Before every paid provider attempt, show the exact plan and receive clear user approval after disclosure. An approval ID is a single-use capability, not proof that a human reviewed the plan. It expires after 30 minutes and is consumed before the attempt. - Never request, print, put on a command line, or write an API key. Plugin configuration supplies it as sensitive user configuration. Standalone scripts read only GEMINI_API_KEY and ignore generic Google key aliases.