Nano Banana Pro (Morfeo)

Generate and edit images using Google's Nano Banana Pro (Gemini 3 Pro Image) API. Use when the user asks to generate, create, edit, modify, change, alter, or update images. Also use when user references an existing image file and asks to modify it in any way (e.g., "modify this image", "change the background", "replace X with Y"). Supports both text-to-image generation and image-to-image editing with configurable resolution (1K default, 2K, or 4K for high resolution). DO NOT read the image file first - use this skill directly with the --input-image parameter.

Security Scan

The skill largely does what its description says (generate/edit images with Gemini), but there are several inconsistencies (missing declared env var, unclear install/dependency instructions, and minor instruction/code mismatches) that should be resolved before trusting it.

Purpose & Capability

The name, description, SKILL.md, and the included script all align with an image-generation/editing skill using Gemini (gemini-3-pro-image-preview). However the registry metadata declares no required environment variables while both SKILL.md and the script require an API key (GEMINI_API_KEY or --api-key). This is an incoherence between claimed requirements and actual runtime needs.

Instruction Scope

The runtime instructions and the script stay within image generation/editing scope: building prompts, sending image bytes to the Google GenAI client, and saving output images. One point to note: the SKILL.md advises 'DO NOT read the image file first' (meaning the agent shouldn't inline image bytes into conversation), whereas the script itself opens input/reference image files locally to include them in the API request — this behavior is expected for the skill but the phrasing could confuse non-technical users or agents.

Install Mechanism

There is no install spec (instruction-only install), which minimizes installer risk. However the script lists dependencies (google-genai and pillow) in a header comment but provides no automated install step — users/agents will need those packages available. This is a documentation/packaging omission rather than an outright risk, but it should be fixed.

Credentials

The script requires an API key (checked via --api-key or GEMINI_API_KEY) and SKILL.md shows endpoints that use API_KEY and (in examples) PROJECT_ID/REGION, yet the skill metadata declares no required env vars or primary credential. That omission is problematic: the skill will fail without the key, and the metadata does not warn users to provide it. Otherwise the requested environment access is proportional (only a service API key) and there are no unrelated credential requests.

Persistence & Privilege

The skill is not always-enabled, does not request persistent system privileges, and does not modify other skills or system-wide settings. It runs as a one-off CLI script that reads local image files and writes output images to the user's working directory.

What to consider before installing

Before installing or running this skill: - Expect to provide a Gemini API key (GEMINI_API_KEY or pass --api-key). The registry metadata currently omits this; verify and supply the key only if you trust the endpoint and code. - The script will open local image files and send image bytes to Google's Generative API — do not use sensitive/private images unless you are comfortable they will be transmitted to Google. - The package lists dependencies (google-genai, pillow) but there is no install step; ensure those Python packages are installed in a controlled environment (virtualenv) before running. - Confirm the source (homepage is none, source unknown). Lack of provenance increases risk; prefer skills with a verifiable source or repository. - If you will deploy this for others, ask the maintainer to: (1) declare GEMINI_API_KEY in the skill metadata, (2) include an install spec or requirements file, and (3) clarify the 'do not read image file first' wording to avoid agent confusion. - General safety: use a least-privilege API key, monitor API usage/billing, and inspect the code locally before running. If you need higher assurance, request a signed/verified source or run in an isolated environment.

Nano Banana Pro Image Generation & Editing

Generate new images or edit existing ones using Google's Nano Banana Pro API (Gemini 3 Pro Image).

API Technical Specification

Endpoints & Authentication

Google AI Studio (Public Preview):

POST https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-image-preview:generateContent?key=${API_KEY}

Vertex AI (Enterprise):

POST https://${REGION}-aiplatform.googleapis.com/v1/projects/${PROJECT_ID}/locations/${REGION}/publishers/google/models/gemini-3-pro-image-preview:predict

Model IDs

• API: gemini-3-pro-image-preview
• SDK interno: nanobanana-pro-001

Parameters

Parameter	Values	Description
aspect_ratio	1:1, 4:3, 3:4, 16:9, 9:16	Output aspect ratio
output_mime_type	image/png, image/jpeg	Output format
reference_images	Array (max 14)	Reference images for consistency
reference_type	CHARACTER, STYLE, SUBJECT	How to use reference
person_generation	ALLOW_ADULT, DONT_ALLOW, FILTER_SENSITIVE	Person generation policy
image_size	1K, 2K, 4K	Output resolution

Reference Types

• STYLE: Transfer visual style, color palette, mood from reference
• CHARACTER: Maintain facial features, traits consistency across images
• SUBJECT: Keep the subject/product consistent (use for product photography!)

Advanced Capabilities

• Text Rendering: Native text rendering without spelling errors
• In-context Editing: Send existing image + modification prompt (automatic in-painting)
• High Resolution: Native upscale to 4K via upscale: true

Usage

Run the script using absolute path (do NOT cd to skill directory first):

Generate new image:

uv run ~/.clawdbot/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "your image description" \
  --filename "output-name.png" \
  [--resolution 1K|2K|4K] \
  [--api-key KEY]

Edit existing image:

uv run ~/.clawdbot/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "editing instructions" \
  --filename "output-name.png" \
  --input-image "path/to/input.png" \
  [--resolution 1K|2K|4K]

With reference image (product/style/character consistency):

uv run ~/.clawdbot/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "your description" \
  --filename "output-name.png" \
  --reference-image "path/to/reference.jpg" \
  --reference-type SUBJECT|STYLE|CHARACTER \
  [--resolution 1K|2K|4K]

Important: Always run from the user's current working directory so images are saved where the user is working, not in the skill directory.

Resolution Options

• 1K (default) - ~1024px resolution
• 2K - ~2048px resolution (recommended for most uses)
• 4K - ~4096px resolution (high quality)

Map user requests:

• No mention → 1K
• "low resolution", "1080", "1080p", "1K" → 1K
• "2K", "2048", "normal", "medium resolution" → 2K
• "high resolution", "high-res", "hi-res", "4K", "ultra" → 4K

API Key

The script checks for API key in this order:

1. --api-key argument
2. GEMINI_API_KEY environment variable

Filename Generation

Format: {timestamp}-{descriptive-name}.png

• Timestamp: yyyy-mm-dd-hh-mm-ss (24-hour format)
• Name: Descriptive lowercase with hyphens

Examples:

• 2025-11-23-14-23-05-japanese-garden.png
• 2025-11-23-15-30-12-sunset-mountains.png

---

Prompt Engineering Framework

You are an expert Prompt Engineer specializing in Nano Banana Pro. Transform basic user ideas and reference images into high-fidelity, descriptive prompts.

1. Input Analysis

When provided with a user idea and reference images, evaluate:

• Subject Matter: Identify primary actors, objects, or focal points
• Reference Utility: Determine if image provides composition (layout), style (aesthetic/texture), or character (specific features)
• Text Requirements: Note any specific text to render within the image

2. Prompt Construction Framework

Structure optimized prompts using this hierarchy:

Core Subject & Action

Clear description of "who" or "what" is doing "what."

Style & Medium

Specify artistic medium:

• Hyper-realistic photography
• Oil painting
• 3D render
• Minimalist vector
• Commercial food photography
• Editorial style

Reference Integration

Explicitly instruct on how to use uploaded images:

"Retain the product packaging from the reference image as the hero element" "Apply the warm lighting aesthetic from Reference A"

Technical Details

Lighting:

• Cinematic rim lighting
• Soft diffused sunlight
• Harsh strobes
• Warm tungsten lighting
• Golden hour warmth

Composition:

• Wide-angle shot
• Macro detail
• Bird's-eye view
• Shallow depth of field
• Product as hero element

Color Theory:

• Monochromatic blue
• High-contrast complementary
• Warm amber tones
• Dark moody palette

Text Rendering: Use double quotes for specific text:

"The word 'FUTURE' written in bold, brushed-metal 3D lettering across the center"

3. Optimization Rules

DO:

• Use descriptive positive language
• Expand vague terms ("cool" → "iridescent", "pretty" → "ethereal", "realistic" → "photorealistic 8k texture")
• Maintain consistency with reference images
• Use powerful adjectives for mood ("gritty," "serene," "industrial," "whimsical")
• Specify "8k texture detail" or "8k photorealistic detail" for quality

DON'T:

• Use negative prompts (say what you want, not what you don't)
• Contradict visual data in reference images
• Use vague terms without expansion

4. Product Photography Best Practices

When generating images with products as protagonists:

1. Always use --reference-type SUBJECT to maintain product consistency
2. Describe the product prominently in the prompt:

   "Milkaut Crematto container with blue label and red lid prominently displayed"

3. Position product as hero element:

   "the product container as co-star product placement" "product container in sharp focus"

4. Include product in scene naturally:

   "positioned next to", "beside", "prominently arranged"

Example Product Photography Prompt:

Hyper-realistic commercial food photography with a [PRODUCT NAME] container 
prominently displayed next to [FOOD ITEM], [food description], 
[setting/background], [lighting style], the [product] as hero element, 
8k photorealistic detail

5. Output Format

Provide the optimized prompt in English, without additional commentary.

---

Examples

Product + Food Scene

uv run ~/.clawdbot/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "Hyper-realistic commercial food photography with a Milkaut Crematto container prominently displayed next to a gourmet double smash burger with perfectly melted cheddar cheese cascading down juicy beef patties, artisan brioche bun, wisps of steam rising, dark moody background with dramatic rim lighting, the cream cheese container as hero product placement, 8k texture detail" \
  --filename "2026-01-28-product-burger.png" \
  --reference-image "product-photo.jpg" \
  --reference-type SUBJECT \
  --resolution 2K

Style Transfer

uv run ~/.clawdbot/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "Using the warm golden hour aesthetic from the reference, create a serene Japanese garden with cherry blossoms, koi pond reflecting soft pink petals, traditional wooden bridge, ethereal morning mist, 8k photorealistic detail" \
  --filename "2026-01-28-japanese-garden.png" \
  --reference-image "style-reference.jpg" \
  --reference-type STYLE \
  --resolution 2K

Image Editing

uv run ~/.clawdbot/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "Change the background to a dramatic sunset over mountains, maintain the subject in sharp focus" \
  --filename "2026-01-28-edited-sunset.png" \
  --input-image "original.jpg" \
  --resolution 2K