ship-learn-next

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

• Third-party content exposure detected (high risk: 0.70). The required runtime workflow reads a user-provided file path (“Read the Content” / “Read the tool to analyze the content”), so if that file contains outsider-authored free text (e.g., a transcript/article/notes the user didn’t author), that text is ingested into the agent’s LLM context for lesson extraction.