echo-jobs

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

• Third-party content exposure detected (high risk: 0.70). The required workflow ingests user-provided inputs like interview transcripts, support ticket themes, NPS verbatims, or issue/PR text, which are often authored by outsiders; at runtime that free-form outsider text is read and then summarized into the agent’s LLM context as “Evidence: [quotes or behaviors from input]”.