product-behavior-signals

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly requires ingesting "recent interview/UGC transcripts, community threads, customer support logs" and uses qualitative quotes in the "Pattern mining" and "Insight synthesis" steps, which indicates the agent will read untrusted, user-generated third-party content (community threads/UGC) that can influence decisions and actions.