multimodal-ai

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

• Third-party content exposure detected (high risk: 0.90). This skill accepts and processes arbitrary external media URLs and user-provided files (e.g., analyzeImageUrl, compareImages, extractStructuredData, processMultiPageDocument, and the UnifiedMultimodalPipeline that takes image URLs or audio files) and sends that untrusted third-party content to vision/transcription models for analysis, so the agent will read/interpret public/user-generated content and could be exposed to indirect prompt injection.