customer-persona

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s documented runtime workflow uses third-party search/answer components (e.g., belt app run tavily/search-assistant and exa/search/exa/answer) on attacker-controlled free-text queries, which can cause the LLM to ingest untrusted text from web/search results without selecting a specific trusted item.