Sandbox
@AgriciDaniel/banana-claude

Claude Code skill for Gemini image workflows

Banana Claude adds a review-first image workflow to Claude Code. It plans the request offline, shows the exact prompt and settings, then uses Gemini to create or edit images after you approve the paid attempt. The repo also includes comparison, continuation, review, and local typesetting steps.

1,032 stars222 forksPythonUpdated 18d ago
Who it's for

Builders who want Claude Code to create and review Gemini images from plain-language prompts.

What it delivers

You can turn a rough visual request into a reviewed image plan before spending a paid Gemini request.

What it does

Offline planning

Builds a visual brief before any API call and shows the prompt, model, data settings, output destination, and estimate.

Single-use approvals

Requires a fresh user approval for one provider attempt, and a changed plan needs a new approval.

Generate, edit, continue, compare, and review

Supports new image generation, one-change edits, session continuation, bounded prompt comparisons, and pixel review.

Privacy-conscious output

Saves the image plus a metadata sidecar while keeping API keys and raw prompts out of public surfaces by default.

Local typesetting

Lets you add approved copy, fonts, logos, and trusted raster art locally when exact text matters.

Upgrade and cleanup path

Includes migrations and a redacted scan for older installs that may have left behind keys, MCP entries, or obsolete skill files.

How to get it

  1. 1Requirements: Claude Code 2.1.199 or newer, Python 3.11 or newer with python3 on PATH,…
    /plugin marketplace add AgriciDaniel/banana-claude
    /plugin install banana-claude@banana-claude-marketplace
    /plugin enable banana-claude@banana-claude-marketplace
    /reload-plugins
  2. 2Then ask for the image you want
    /banana-claude:banana generate an urban 16:9 GitHub hero with clean left-side copy space

README

Banana Claude character cycling through thirty artistic urban looks while the Banana Claude title and pixel banana remain fixed

Banana Claude

Create, edit, compare, and review Gemini images from Claude Code.

Validation Version 3.0.0 MIT License

Banana Claude turns a plain-language creative request into a planned and reviewable Gemini image workflow. It shows the prompt, model, data settings, output destination, and nominal estimate before any paid request.

Here's the skill, here's what it does, here's how to install it.

Start here

Requirements: Claude Code 2.1.199 or newer, Python 3.11 or newer with python3 on PATH, and a Gemini API key for a billing-enabled Google AI project.

/plugin marketplace add AgriciDaniel/banana-claude
/plugin install banana-claude@banana-claude-marketplace
/plugin enable banana-claude@banana-claude-marketplace
/reload-plugins

Then ask for the image you want:

/banana-claude:banana generate an urban 16:9 GitHub hero with clean left-side copy space

Banana Claude plans offline first. Review the exact plan, then approve the paid request when it matches your intent.

The workflow is simple

  1. Ask: describe the outcome in normal language.
  2. Plan: Banana Claude builds the visual brief and chooses a route.
  3. Review: inspect the prompt, references, privacy settings, and estimate.
  4. Approve: one short-lived approval authorizes one provider attempt.
  5. Create: the image and a privacy-conscious metadata sidecar are saved.
  6. Check: review the actual pixels, then fix, regenerate, or ship.

What you can make

  • Generate: campaign visuals, covers, product scenes, diagrams, concepts, social assets, and more.
  • Edit: make one clear change while protecting identity, geometry, and brand details.
  • Continue: iterate with stored Flash or Pro sessions, or attach the last result as a fresh reference.
  • Compare: test up to three prompts across three model routes in one bounded portfolio.
  • Review: check copy, crop, composition, consistency, artifacts, rights, and provenance.
  • Typeset: add approved copy, fonts, logos, and trusted raster art locally when exact text matters.

Banana Claude routes across Google's current Nano Banana 2 Lite, Nano Banana 2, and Nano Banana Pro models. See the dated Gemini model reference for current capabilities, pricing, and the deprecated compatibility route.

Built for control

  • The plugin installs disabled because image generation is a paid service.
  • API keys stay out of commands, request URLs, metadata, and public CI.
  • Every paid tool shows its plan and requires a user decision before execution.
  • One approval permits one provider attempt. A changed plan needs a new approval.
  • Cost records stay private and omit raw prompts by default.
  • Generated output is never treated as automatically correct or production ready.

Gemini output is probabilistic. Banana Claude cannot guarantee consistency, spelling, geometry, policy acceptance, rights clearance, or a final invoice. Those remain review conditions, not marketing claims.

Read Security and privacy before production use.

Upgrade from public 1.4.1 or 2.1.0 installs

Do not simply overwrite an older public install. It may have left an unpinned third-party MCP entry, a raw Google key in Claude settings, or an obsolete skill directory. Version 3 includes a redacted scan, review-first cleanup, and explicit state migrations.

Follow the upgrade guide before enabling version 3. Rotate any key that an older setup stored on disk or exposed in shell history.

Documentation

Contributing

Contributions are welcome. Read CONTRIBUTING.md before opening a pull request. For help or a reproducible bug, open an issue without keys, private media, or full environment dumps.

License

MIT License, see LICENSE.

Built by Agrici Daniel, AI Workflow Architect.

Blog · AI Marketing Hub · YouTube · More open-source projects

Files in the repo

Repository payload23 top-level entries
  • .claude-plugin
  • .github
  • agents
  • docs
  • screenshots
  • skills
  • tests
  • tools
  • .gitattributes
  • .gitignore
  • .mcp.json
  • CHANGELOG.md
  • CITATION.cff
  • CLAUDE.md
  • CODE_OF_CONDUCT.md
  • CODEOWNERS
  • CONTRIBUTING.md
  • install.sh
  • LICENSE
  • pyproject.toml
  • README.md
  • requirements-dev.txt
  • SECURITY.md

Discussion (0)

Ask about usage, or say what you built with it

Sign in to join the discussion.

No comments yet. Be the first to say what this is good for.

More skills

Vincentwei1021/
anything2explainer

Topic in, narrated explainer video out. A Claude Code / Codex skill that turns any topic into a black-canvas motion-graphics explainer video with TTS voiceover, subtitles and a chapter progress bar. Chinese or English; every frame drawn in code with Remotion.

666

Taste-Skill - gives your AI good taste. stops the AI from generating boring, generic slop

86k

Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self-contained HTML with motion and crisp export.

57k
img2threejs/
img2threejs

Rebuild the object in a reference image as a code-only, procedural, quality-gated, animation-ready Three.js model. Token-efficient image-to-3D.

16k