biological-expert

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill provides utility classes for molecular biology and bioinformatics without any detected malicious patterns or risky operations. All external resources provided are established scientific institutions and databases.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface that is common for scientific analysis tools, though no specific exploits were identified.
  • Ingestion points: The DNAAnalyzer class (sequence parameter) and RNASeqAnalyzer class (counts_matrix parameter) ingest potentially untrusted scientific data strings.
  • Boundary markers: The implementation lacks explicit delimiters or warnings to ignore instructions that might be embedded in the biological data strings.
  • Capability inventory: The skill has access to Bash(python:*), Read, Write, and Edit tools, allowing it to execute code and modify files based on its analysis.
  • Sanitization: Inputs are normalized (e.g., .upper()) but do not undergo strict validation against non-biological instruction-like patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 02:46 PM
Security Audit — agent-trust-hub — biological-expert