bio-population-genetics-linkage-disequilibrium

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • Standard Command Execution (SAFE): The skill utilizes shell commands to interact with PLINK and VCFtools. All commands are standard bioinformatics workflows for calculating LD, pruning variants, and clumping GWAS results.
  • Trusted Tooling (SAFE): The recommended installation methods use official package managers (conda, pip) and point to reputable sources (bioconda, PyPI) for well-known scientific software.
  • Python Implementation (SAFE): Python code snippets use established libraries (scikit-allel, pandas, matplotlib) for data processing and visualization without any dynamic execution or network access beyond local file reading.
  • Indirect Prompt Injection Surface (LOW): As a data processing skill, it ingests external genomic files (VCF, PLINK BED/BIM/FAM). While these represent ingestion points for untrusted data, the structured nature of these binary and text formats makes them highly resistant to prompt injection attacks against the agent.
  • No Persistence or Exfiltration (SAFE): There are no attempts to modify system configurations, establish persistence, or transmit data to external servers.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:35 PM