biomedical-skills

Pass

Audited by Gen Agent Trust Hub on Apr 5, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill set references and provides examples for downloading biological data and tools from well-known scientific services. For example, it includes instructions for fetching genome data from NCBI, nucleotide sequences from the European Nucleotide Archive (ENA), and the AlphaFold protein structure database. It also provides the download link for the official NCBI Datasets CLI tool.
  • [INDIRECT_PROMPT_INJECTION]: Several skills (e.g., bioagents-guide, med-researcher-r1-guide, genomas-guide) describe agentic systems that ingest data from external sources such as PubMed, ClinicalTrials.gov, and local genomic files.
  • Ingestion points: External text data from literature databases and structured biological data files.
  • Boundary markers: None explicitly defined in the conceptual prompts.
  • Capability inventory: Includes execution of bioinformatics tools (STAR, GATK, DESeq2) and network requests to scientific APIs.
  • Sanitization: The guides do not explicitly detail sanitization steps for the ingested text, focusing instead on clinical extraction (PICO framework). Given the primary purpose of biomedical research, the risk is negligible.
  • [COMMAND_EXECUTION]: The guides provide educational snippets for running standard bioinformatics commands (e.g., fastqc, plink2, samtools) and installing specialized research packages (e.g., genomas, scispacy). These commands are standard for the described research tasks and operate on local data.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 5, 2026, 11:14 PM