competitor-teardown

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

• Third-party content exposure detected (medium risk: 0.30). The skill runtime path explicitly uses free-text outsider queries to perform web/search and page extraction (e.g., belt app run tavily/search-assistant and belt app run tavily/extract plus infsh/agent-browser with attacker-chosen url/target), meaning the LLM ingests outsider-authored content from those fetched sources.