tracking-taxonomy-updates

Pass

Audited by Gen Agent Trust Hub on Oct 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references and provides instructions for downloading authoritative taxonomy data from established scientific sources, including the NCBI FTP server (taxdump) and the GTDB portal. These are well-known scientific repositories and the downloads are for standard reference data for scientific analysis.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes sequence metadata from quickclade.tsv which is then used by the agent to determine the domain-routing workflow and generate reports. This ingestion of external data creates a surface where malicious content could attempt to influence the agent's instructions or subsequent outputs. Evidence: Ingestion point is quickclade.tsv (processed by scripts/quickclade_to_routing.py); boundary markers are absent; capabilities include tool execution via sbatch and report generation; sanitization is limited to basic string manipulation.
  • [COMMAND_EXECUTION]: The skill uses sbatch to submit jobs to a SLURM scheduler and docker/apptainer to run containers for bioinformatics analysis. These tools are used for their intended scientific purpose, and the execution scripts in scripts/submit_taxonomy.sh include validation for tool names and use absolute paths to mitigate command injection risks during standard operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 5, 2026, 02:46 PM