ai-sdk-ui

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

• Third-party content exposure detected (high risk: 0.70). The skill ingests untrusted user-generated content that the agent will read — e.g., the importChat / export/import conversation example in streaming-patterns.md and message-persistence.tsx (setMessages from an uploaded JSON) and the experimental_attachments flow in use-chat-attachments.tsx, which accept arbitrary user files/contents that the assistant will process, enabling indirect prompt injection.