banana-split

banana-split is a command for Claude Code from seancrowe01/ads-machine. It costs 44 tokens per session (1,511 once invoked), scanned A, original, MIT.

An image-generation command that uses Google's Gemini CLI and Nano Banana models to create visual assets from descriptions.

In plain words
What is it for?
Use it to make advertisements, social-media images, thumbnails, logos, hero images, and other visual assets saved in the project's creatives folder.
Why use it?
It gives developers a defined way to produce images for marketing and product work, including a fallback when the Gemini CLI is unavailable.

Command for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/seancrowe01/ads-machine/banana-split
Clone the repo
git clone --depth 1 https://github.com/seancrowe01/ads-machine

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for banana-split

README.md
[![agentmods](https://agentmods.dev/badge/commands/seancrowe01/ads-machine/banana-split.svg)](https://agentmods.dev/commands/seancrowe01/ads-machine/banana-split)
Your own site
<a href="https://agentmods.dev/commands/seancrowe01/ads-machine/banana-split"><img src="https://agentmods.dev/badge/commands/seancrowe01/ads-machine/banana-split.svg" alt="Measured on agentmods" height="20"></a>
Per session 44 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,511 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00044 $0.01511
Opus 5 $0.00022 $0.00756
Sonnet 5 $0.00009 $0.00302
Haiku 4.5 $0.00004 $0.00151

Measured 4d ago against content hash 8e615e859d49, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

banana-split scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

curl -s https://api.openai.com/v1/images/generations \
.claude/commands/banana-split.md · 211 lines

How it starts

The opening of the file, as written. The whole thing — 211 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Banana Split — Image Generation

You are a creative director and image generator. You take a description, brief, or Pipeline record and produce high-quality images using Nano Banana models via Gemini CLI.

What you produce: Generated images saved to creatives/ folder with text overlay specs when needed.


Models

Model Best For Command
Nano Banana 2 Fast drafts, iterations, social content gemini --yolo "/generate 'prompt'"
Nano Banana Pro High quality, ad creatives, hero images gemini --yolo "/generate with pro model: 'prompt'"

Default: Nano Banana 2 for speed. Use Pro when the user asks for higher quality or when generating final ad creatives.


Setup

Requires Gemini CLI installed:

npm install -g @google/gemini-cli
gemini  # First run: authenticate with Google account

If Gemini CLI is not installed, fall back to:

  1. DALL-E via OpenAI API (OPENAI_API_KEY in .env)
  2. Flux via fal.ai (FAL_KEY in .env)
  3. Output the prompt only -- user pastes into their preferred tool

How It Works

Quick Generate

User says what they want. You generate it.

User: "generate a hero image for a boxing gym landing page"
You: Build the prompt, run Gemini, save the output

From Ad Pipeline

If working from a Pipeline record, pull the brief:

Use Airtable MCP: list_records
  filter: {Status}='Scripted' OR {Status}='Briefed'
  fields: Name, Angle, Format, Hook, Primary Text, Headline

Use the ad details to inform the image direction.


Prompt Building

Structure

Every prompt follows this format:

[SUBJECT]: What is in the image
[STYLE]: Photography style or illustration style
[COMPOSITION]: Where elements sit, camera angle
[MOOD]: Emotional tone
[DETAILS]: Specific requirements (aspect ratio, text space, etc.)

Ad Creative Rules

When generating for ads (user mentions "ad", "creative", "campaign", or is working from Pipeline):

  • Always leave clear space for text overlay (top third or bottom third)
  • No text in the generated image -- AI text looks terrible. Specify overlay separately.
  • Match the angle to the visual:

Read the full file on GitHub · 211 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 211 lines · 44 tokens per session scan A 8e615e859d49

Subscribe to this mod's changes

banana-split is a command published in the GitHub repository seancrowe01/ads-machine (20 stars, last pushed 5mo ago), licensed MIT. It adds 44 tokens to every session and 1,511 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.