nano-banana

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill explicitly accepts external image URLs in the images input (e.g., "images": ["https://example.com/landscape.jpg"]) and provides an enable_google_search grounding option in the Input Options and examples, showing it fetches and ingests public web/search content which the model uses to influence generation—exposing the agent to untrusted third-party content that could carry indirect instructions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill depends on the inference.sh platform (https://inference.sh) and its infsh CLI at runtime to run remote apps like google/gemini-3-pro-image-preview, which executes remote model code and thereby relies on externally-hosted code to control generation.