synthesis

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

• Third-party content exposure detected (high risk: 0.75). The skill’s required workflow ingests “raw inputs” such as interview notes/transcripts and other qualitative material that may include outsider-authored free text (e.g., customer interview transcripts or comments), which the agent would read and then incorporate into the LLM context during clustering/evidence stacking.