nih

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

• Third-party content exposure detected (low risk: 0.10). The agent searches and reads PubMed articles from nih.gov via search tools, which is an external content search/fetch source where authors can publish text.

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

• Potentially malicious external URL detected (medium risk: 0.30). The skill configures a remote Model Context Protocol (MCP) server from unbrowse.ai (https://unbrowse.ai/mcp/nih.gov), which provides dynamic operational tools and code execution behavior at runtime from an unofficial third-party publisher, representing an unverifiable runtime external dependency.