nano-banana-pro
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text inputs to generate images via the Gemini API, creating an attack surface where malicious instructions could be embedded in prompts.
- Ingestion points: Untrusted data enters the agent context through the
contentsparameter in Python/cURL examples and thepromptvariable in the Next.js API route withinSKILL.md. - Boundary markers: The instructions lack explicit delimiters or system instructions to the model to ignore potentially malicious commands embedded within user prompts.
- Capability inventory: The skill demonstrates the capability to perform network operations to external APIs and write image data to the local file system (e.g.,
output.pnginSKILL.md). - Sanitization: The provided examples do not include input validation, escaping, or filtering for user-provided prompts.
- [EXTERNAL_DOWNLOADS]: The skill facilitates communication with Google's official Generative Language API at
generativelanguage.googleapis.comto fetch image and text data. - [COMMAND_EXECUTION]: Documentation provides examples for using the
curlutility to send requests to Google's REST API endpoints.
Audit Metadata