bio-phylogenomics

Pass

Audited by Gen Agent Trust Hub on Feb 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • Indirect Prompt Injection (LOW): The skill ingests untrusted biological sequence data which could contain malicious instructions targeting the agent logic. Ingestion points: markers.faa, alignments.fasta. Capability inventory: Execution of shell commands (iqtree, VeryFastTree) and Python scripts. Sanitization: Not specified.
  • Dynamic Execution (LOW): Documentation provides templates for generating and executing Python scripts using shell HEREDOCs (python << EOF) for tree manipulation.
  • External Downloads (LOW): The skill recommends installing software via pip and conda from external repositories. Manual installation instructions for iqtree use an insecure HTTP link.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 19, 2026, 05:19 PM