geo-database
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process datasets from the Gene Expression Omnibus (GEO), which contains content submitted by the scientific community. This creates a potential surface for indirect prompt injection.
- Ingestion points: Untrusted data enters the agent's context through metadata retrieval in SKILL.md (via GEOparse.get_GEO and Entrez.esummary) and full records in references/geo_reference.md.
- Boundary markers: No specific boundary markers or instructions to ignore embedded prompts are implemented when parsing or displaying retrieved metadata like 'Series Summary' or 'Sample Characteristics'.
- Capability inventory: The skill possesses capabilities to write files to the local system (pd.to_csv, download_supplementary_files), perform network operations (ftplib.FTP, Bio.Entrez), and generate visualizations.
- Sanitization: There is no evidence of sanitization or filtering applied to external content before it is interpolated into the agent's context or stored locally.
- [EXTERNAL_DOWNLOADS]: The skill facilitates the download of configuration, metadata, and large biological datasets from NCBI servers.
- Evidence: Downloads occur from ftp.ncbi.nlm.nih.gov and eutils.ncbi.nlm.nih.gov. These are well-known services provided by the National Center for Biotechnology Information.
Audit Metadata