nano-banana
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
uvto manage its runtime environment, fetching standard Python packages includinggoogle-genai(Google's official SDK) andpillow(a well-known imaging library). These are recognized as safe, established dependencies from trusted organizations. - [INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection as it processes user-provided prompts and local reference images which are then passed to an external image generation API.
- Ingestion points: User input provided via the
--promptargument and image files specified via the--referenceparameter inscripts/image.py. - Boundary markers: The skill does not implement specific boundary markers or 'ignore' instructions when prepending aspect ratio metadata to the user prompt.
- Capability inventory: The skill possesses network access capabilities (to communicate with Google's Gemini API) and file-system write permissions (to save the generated images).
- Sanitization: The skill relies on the
google-genaiclient library for data transmission and does not perform manual sanitization of the input strings, which is standard for generative tasks.
Audit Metadata