bgpt-paper-search

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

• Third-party content exposure detected (high risk: 0.85). The required workflow calls a remote MCP server (bgpt.pro) via the search_papers tool, and the LLM will ingest the server’s returned structured text extracted from full-text scientific studies authored by outsiders (i.e., not the operating user’s own content).

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

• Potentially malicious external URL detected (high risk: 0.70). The MCP configuration runs npx/mcp-remote connecting to https://bgpt.pro/mcp/sse (and offers an npx bgpt-mcp alternative), so at runtime this external server/package is fetched/executed and its tool responses directly affect agent behavior via the remote MCP tool interface.