bgpt-paper-search
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documentation provides instructions to configure a remote MCP server using npx mcp-remote and npx bgpt-mcp, which involve downloading code from the npm registry and connecting to the bgpt.pro service endpoint.
- [INDIRECT_PROMPT_INJECTION]: The skill processes structured experimental data extracted from full-text papers via an external API.
- Ingestion points: Data returned by the search_papers tool, including study methods, results, and conclusions.
- Boundary markers: No explicit delimiters are used to isolate retrieved study data from agent instructions.
- Capability inventory: The agent uses retrieved paper data to answer queries, build evidence tables, and perform synthesis.
- Sanitization: No sanitization or validation of the retrieved scientific content is described in the skill instructions.
Audit Metadata