cro-methodology

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md itself is static documentation, but it explicitly instructs the runtime process to ingest “exit surveys, live chat logs, support tickets, sales calls, and reviews” (i.e., outsider-authored free-text from customers/competitors) into the agent’s LLM context for objection/research mining.