keyword-clustering

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

• Third-party content exposure detected (medium risk: 0.30). In the runtime workflow, the agent reads Search Console data via get_search_console_performance (real queries and pages) when connected, and those query/page strings can include outsider-authored user-generated text (the search queries entered by users), meaning the LLM ingests free text without selecting a specific item first.