tavily-research

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly says the tool does "automatic source gathering" and "comprehensive research grounded in web data," and scripts/research.sh calls the Tavily MCP endpoint (https://mcp.tavily.com/mcp) invoking the "tavily_research" tool, which indicates the agent ingests and synthesizes open/public web content that could contain untrusted, user-generated instructions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill invokes npx -y mcp-remote https://mcp.tavily.com/mcp at runtime to perform OAuth, which will fetch and execute remote npm package code (and also calls https://mcp.tavily.com/mcp via curl), meaning remote content is executed during runtime and is required for the authentication flow.