nano-banana

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

• Third-party content exposure detected (high risk: 0.80). The skill's documentation explicitly shows a "Google Search Grounding" example in SKILL.md (Advanced Features) using tools=[{"google_search": {}}], which would fetch real-time public web results (untrusted third‑party content) that the model would read to generate images, enabling indirect prompt injection.