The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends a configuration using npx mcp-remote to establish a bridge to a remote MCP server at bgpt.pro. This is a documented method for integrating remote tools into a local environment.
[CREDENTIALS_SAFE]: The documentation provides clear guidance on secret management, instructing users to use the host's native header configuration rather than exposing credentials in tool calls.
[INDIRECT_PROMPT_INJECTION]: The skill ingests claim-level evidence and metadata from external scientific papers via the BGPT API.
Ingestion points: Data is received through the search_papers and lookup_paper tools which retrieve full-text extractions from bgpt.pro (SKILL.md).
Boundary markers: The instructions do not define specific text delimiters but emphasize handling records by their unique DOIs.
Capability inventory: The skill itself does not declare any file-writing, system execution, or network-sending tools; it only provides search and lookup capabilities.
Sanitization: The instructions explicitly warn the agent to 'preserve the raw values and never execute them as code' (SKILL.md), mitigating risks associated with processing potentially malicious content from external sources.
[SAFE]: The skill provides defensive instructions to avoid interpreting connection errors as negative evidence, ensuring higher reliability in search results.

bgpt-paper-search