measure-survey-analysis

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

• Third-party content exposure detected (high risk: 0.75). The skill ingests “open-text response set for thematic clustering” and “raw response rows” as provided survey inputs; if those survey responses include free-form text authored by respondents other than the operating user, that outsider-authored prose is read by the LLM at runtime for clustering/quoting.