ZenMux Image Gen (Nano Banana 2)

Generate images from text prompts using ZenMux API (Vertex AI protocol with Gemini models). Use when: (1) User wants to generate/create images from text desc...

Security Scan

The skill generally behaves like an image-generation wrapper for ZenMux, but packaging and metadata inconsistencies (missing declared env var) plus a truncated script file reduce confidence and warrant caution before installing.

Purpose & Capability

The name, description, SKILL.md, and the script all target ZenMux image generation (Vertex AI/Gemini models). That purpose aligns with the code's network calls to https://zenmux.ai/... and the CLI options. However, the registry metadata claims 'Required env vars: none' whereas SKILL.md and the script explicitly require ZENMUX_API_KEY — this mismatch suggests sloppy metadata or incomplete publishing.

Instruction Scope

Runtime instructions and the script stay within the stated purpose: building a request from a prompt (and optional input image), POSTing to the ZenMux Vertex-like endpoint, and saving returned image bytes. The SKILL.md warns not to pass API keys on the command line (good). The SKILL.md and script do not request unrelated system files, other credentials, or external endpoints beyond zenmux.ai.

Install Mechanism

This is an instruction-only skill with no install spec; the included script is executed directly. No download/install steps or third-party package pulls are present in the manifest, which reduces install-time risk.

Credentials

The code and SKILL.md require a ZENMUX_API_KEY (or --api-key) to call the API, which is proportionate to the task. But the registry metadata declares no required env vars — an inconsistent declaration. Additionally, the script provides a --api-key CLI flag even though the SKILL.md explicitly warns against using it (exposure risk via shell history). No other credentials are requested, which is appropriate.

Persistence & Privilege

The skill does not request persistent/always-on presence and does not modify other skills or system config. It runs as a normal user-invokable script and does not escalate privileges in the visible code.

What to consider before installing

This skill appears to implement a legitimate ZenMux image-generation client, but exercise caution: (1) the package metadata does not declare the ZENMUX_API_KEY even though both SKILL.md and the script require it — verify the publisher and why metadata is incomplete; (2) the provided script appears truncated in the bundle listing (file output was cut off), so request or review the full script before running — the missing tail could contain unexpected behavior; (3) follow the SKILL.md advice and do not pass API keys on the command line; set ZENMUX_API_KEY in a secure environment variable instead; (4) confirm network endpoints (the script calls only zenmux.ai) and, if you must run it, do so in an isolated environment/container and inspect the full source; (5) if you need higher assurance, ask the publisher for complete source and a signed release or prefer an official client from zenmux.ai.

ZenMux Image Generation

Generate images from text prompts using ZenMux API with Google Gemini models.

Quick Start

# Generate image (requires ZENMUX_API_KEY env var)
python3 scripts/generate_image.py "A cute cat" output.png

# With aspect ratio and resolution
python3 scripts/generate_image.py "A cute cat" output.png --aspect-ratio 16:9 --image-size 2K

API Key Setup

Get your API key from https://zenmux.ai.

# Set environment variable
export ZENMUX_API_KEY="sk-..."

Security Rules:

• NEVER use --api-key flag (visible in shell history)
• NEVER commit API keys to git

Script Options

Option	Description
--model MODEL	Model name (default: google/gemini-3.1-flash-image-preview)
--input-image PATH	Input image for image-to-image generation
--temperature N	Randomness 0.0-2.0 (default: 1.0)
--max-tokens N	Max output tokens
--aspect-ratio RATIO	1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9, 21:9
--image-size SIZE	1K (default), 2K, 4K

Image-to-Image

Edit existing images with text prompts:

python3 scripts/generate_image.py "Make it nighttime" result.png --input-image photo.jpg

Available Models

Model	Alias	Use Case
google/gemini-3.1-flash-image-preview	Nano Banana 2	Fast & quality (default)
google/gemini-3-pro-image-preview	Nano Banana Pro	Best quality
google/gemini-2.5-flash-image	Nano Banana	Faster generation

Prompt Tips

Writing effective prompts:

• Be specific: "A golden retriever puppy playing on grass" instead of "a dog"
• Include details: lighting (golden hour, soft light), style (photorealistic, anime, oil painting)
• For image-to-image: describe the desired change clearly, e.g., "Change to nighttime with city lights"
• Try 10+ words for better results

Troubleshooting

Error	Fix
Input image not found	Check file path; verify file exists
Image too large	Max input size is 20 MB; compress before uploading
401 Unauthorized	Check ZENMUX_API_KEY is set correctly
429 Rate limited	Wait and retry, or upgrade to paid tier
Request timeout	API may be slow; wait and retry
Network error	Check internet connection
No image in response	Try a different prompt or model
Poor image quality	Use more descriptive prompt (10+ words)

Script Features

• Timeout protection: Requests timeout after 300 seconds (5 minutes)
• Auto-create directories: Output directories are created automatically
• Auto-rename on conflict: If output file exists, automatically appends number (e.g., image_1.png, image_2.png)
• Input validation: Checks image format, size, and existence
• Clear error messages: Human-readable error explanations
• Format support: JPG, PNG, WebP, GIF input formats

Batch Generation

Generate multiple images by running the script multiple times with different outputs:

# Generate variations of the same prompt
for i in 1 2 3; do
  python3 scripts/generate_image.py "A cute cat in different poses" "cat_$i.png"
done