firebase-ai-logic

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

• Third-party content exposure detected (high risk: 0.80). The skill's Step 4 "Multimodal (image + text)" example shows analyzeImage(imageUrl) using fetchImageAsBase64(imageUrl) and passing arbitrary image URLs into model.generateContent, which clearly ingests untrusted third‑party content from public URLs that the agent will interpret and that can influence downstream behavior.