evolink-nano-banana-2

What This Skill Does

Nano Banana 2 is a streamlined, high-performance AI image generation skill designed specifically for the OpenClaw agent ecosystem. Powered by the Google Gemini 3.1 Flash image model and facilitated through the unified Evolink API, this skill allows users to bridge the gap between creative thought and visual execution with minimal latency. It provides a robust interface for text-to-image generation and advanced image editing, allowing users to leverage Google’s cutting-edge generative technology within their existing development or creative workflow. Whether you are creating mockups, assets for projects, or experimental visuals, Nano Banana 2 ensures that the process is fast, reliable, and deeply integrated into your agent's capabilities.

Installation

To begin using the skill, ensure you have the mcporter utility or your preferred MCP client (Claude Desktop/Cursor) configured. Use the following command to integrate the media server:

claude mcp add evolink-media -e EVOLINK_API_KEY=your-key -- npx -y @evolinkai/evolink-media@latest

Once the server is configured, run the installation via OpenClaw: clawhub install openclaw/skills/skills/bethune89/evolink-nano-banana-2-1 After installation, verify your API key access by requesting list_models to ensure the agent can communicate with the Evolink gateway.

Use Cases

Nano Banana 2 excels in environments where speed and high-quality inference are required. Use it to generate rapid prototypes for UI/UX design, create stylized concept art for creative writing sessions, or edit existing assets via image-to-image workflows. Because the skill handles async task polling, it is ideal for generating batches of variations based on a single core prompt.

Example Prompts

1. "Generate a minimalist, high-resolution background featuring a geometric interpretation of a nebula in deep blues and purples."
2. "I'm working on a character design. Can you take this uploaded character sketch and generate a 3D-rendered version with a cyberpunk aesthetic?"
3. "Show me the status of my previous image generation task and then list all current files in my cloud storage so I can clear out old samples."

Tips & Limitations

To maximize results, always start by describing the composition and color palette before mentioning specific style references. The model performs best with descriptive, clear prompts rather than long, abstract paragraphs. Remember that uploaded files are temporary, expiring after 72 hours, so ensure you download or save important assets locally. Because images are generated asynchronously, use the check_task tool if the agent does not immediately return a result URL upon request.