ai-provider-mistral-sdk

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly ingests untrusted third-party content via image URLs (see SKILL.md Pattern 7: Vision and examples/embeddings-vision.md's analyzeImageUrl which passes arbitrary imageUrl values like "https://example.com/photo.jpg" to client.chat.complete), causing the model to fetch/interpret external images as part of its workflow and thus enabling indirect prompt injection.