bio-pathway-reactome

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • Indirect Prompt Injection (LOW): The skill processes external data from a CSV file (de_results.csv) which could theoretically contain malicious instructions.
  • Ingestion points: read.csv('de_results.csv') in the 'Prepare Gene List' section.
  • Boundary markers: None present in the code snippets.
  • Capability inventory: File system read/write (read.csv, write.csv), R environment code execution, and potential browser interaction via viewPathway.
  • Sanitization: No explicit sanitization or validation of the input CSV content is shown before processing.
  • External Dependencies (SAFE): The skill relies on Bioconductor packages (ReactomePA, clusterProfiler, enrichplot) and organism-specific annotation databases. While these are external, they are standard libraries within the bioinformatics domain.
  • Data Handling (SAFE): The skill performs local data processing. No evidence of credential exposure or exfiltration to unauthorized external domains was found.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:33 PM