ingest

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

• Third-party content exposure detected (medium risk: 0.65). In doc-import mode, the runtime path extracted from the user’s argument is verified and then the skill reads the full file content (Read the full file content as input payload), which is then embedded into the created source page and thus fed as outsider-authored free text into the agent’s LLM context if the imported file is not authored by the operating user/organization.