grounding-with-google-genai

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

• Third-party content exposure detected (medium risk: 0.30). The runtime path scripts/google_genai_grounding.py → build_interaction_request() (modes search/maps/url) → client.interactions.create(**request) ingests outsider-authored free text via the user-supplied prompt and, in url mode, retrieves and reads outsider-authored contents from user-provided public URLs, as well as search results that are driven by the user query.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill accepts user-provided --url values (e.g., https://example.com/one and https://example.com/two) and includes them in the interaction input and requests the GenAI "url_context" tool at runtime (scripts/google_genai_grounding.py), causing remote page content to be retrieved and directly control the model's context.