nano-banana
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill refers to external resources and tools from the inference.sh platform, including installation instructions hosted on GitHub (github.com/inference-sh) and the belt CLI tool.
- [COMMAND_EXECUTION]: The skill uses the 'belt' command-line utility to interact with the Google Gemini API. These commands are restricted by the 'allowed-tools' field to the 'belt' executable.
- [PROMPT_INJECTION]: The skill accepts user-defined prompts and image URLs which are interpolated into JSON payloads for the 'belt' CLI. This creates an indirect prompt injection surface where malicious inputs could theoretically attempt to influence the image generation model's output.
- Ingestion points: User-provided 'prompt' and 'images' parameters in the 'belt app run' command.
- Boundary markers: Inputs are encapsulated within a JSON string ('--input').
- Capability inventory: Shell execution of the 'belt' CLI tool (SKILL.md).
- Sanitization: No explicit sanitization or escaping logic is defined within the skill instructions, relying on the agent's default handling of JSON data.
Audit Metadata