scientific-db-pubmed-database
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external biomedical data from the PubMed database, creating a potential surface for indirect prompt injection through search results.
- Ingestion points: Data retrieved from NCBI E-utilities endpoints such as
esearch.fcgiandefetch.fcgi(SKILL.md). - Boundary markers: The skill does not explicitly define delimiters for the retrieved literature content.
- Capability inventory: Performs HTTP GET requests using the
requestslibrary to fetch data (SKILL.md). - Sanitization: Uses standard JSON parsing for API responses, which mitigates basic structural injection.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill performs legitimate network operations to the well-known NCBI E-utilities service.
- Evidence: The Python implementation in
SKILL.mdtargetshttps://eutils.ncbi.nlm.nih.gov/entrez/eutilsfor search queries. - Best Practices: Correctley advises storing sensitive credentials like
NCBI_API_KEYandNCBI_EMAILin environment variables rather than hardcoding them.
Audit Metadata