geo-database

Pass

Audited by Gen Agent Trust Hub on Nov 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of large-scale genomic datasets (SOFT, Matrix, and XML formats) from the NCBI GEO repository for analysis. While NCBI is a highly trusted source, processing external data that is then used in analysis workflows or summarized by the agent presents an indirect prompt injection surface.
  • Ingestion points: Data is fetched via Bio.Entrez and the GEOparse library from NCBI domains (eutils.ncbi.nlm.nih.gov and ftp.ncbi.nlm.nih.gov).
  • Boundary markers: No specific delimiters or instructions to ignore embedded content are provided for the fetched biological data.
  • Capability inventory: The skill includes capabilities to write files to the local system using pandas.to_csv(), matplotlib.pyplot.savefig(), and standard file I/O operations in SKILL.md.
  • Sanitization: The provided examples focus on data analysis and do not include explicit sanitization or validation of the remote content beyond standard library parsing.
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to download genomic data and metadata from well-known scientific services.
  • Evidence: References to ftp.ncbi.nlm.nih.gov and eutils.ncbi.nlm.nih.gov for retrieving research datasets. These are established repositories for the scientific community.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 6, 2026, 12:15 PM