docsumo

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly describes actions like "Upload Document from URL" and retrieving "Get Extracted Data"/"Get Document Details" via the Membrane actions, which ingest arbitrary public URLs/documents (third-party/untrusted content) that the agent is expected to read and which can influence subsequent actions (e.g., update-review-status), so indirect prompt injection is possible.