nanobanana

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly supports "Google Search grounding" (shown in SKILL.md and in generate.py / scripts with the --search flag and generate.py setting config_kwargs["tools"] = [{"google_search": {}}]), which fetches open-web search results (untrusted third-party content) and feeds them into the model's generation workflow, enabling indirect prompt injection.