tavily

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and ingests arbitrary public web content (e.g., Tavily Extract /extract with urls, Tavily Crawl/Map with url and instructions, and Tavily Search's include_raw_content) so untrusted third‑party pages would be read and can materially influence the agent's outputs and actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The Tavily Extract/Map/Crawl endpoints fetch user-supplied URLs at runtime (e.g., https://example.com/article1) and can return cleaned/raw page content that is injected into the agent/LLM context to drive answers, meaning fetched external content can directly control prompts.