paperbanana

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow can ingest outsider-authored free text via user-supplied --input/--context (e.g., method.txt, paper.pdf, or --caption) into the LLM’s prompt as the “source context” for planning/critique, which is not guaranteed to be authored by the operating user.