ncbi-sequence-fetch
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests biological data and metadata from the external NCBI E-utilities API, which presents a surface for indirect prompt injection attacks.
- Ingestion points: Data is fetched from the official NCBI API at
https://eutils.ncbi.nlm.nih.gov/entrez/eutilsusing thescripts/ncbi_fetch.pyscript. - Boundary markers: The script returns structured data but does not wrap the variable content in explicit delimiters or include specific instructions for the agent to ignore potential commands embedded within biological text fields (like gene definitions or abstracts).
- Capability inventory: The skill possesses the ability to fetch large volumes of sequence data, perform multi-database searches, and write output to the local file system using the
--outputargument or when creating its internal license tracking file. - Sanitization: While the script correctly parses XML and JSON formats, it does not sanitize or escape the content of text fields before providing them to the agent's context, leaving open the possibility that a malicious record could influence agent behavior.
Audit Metadata