biopython

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill facilitates communication with the National Center for Biotechnology Information (NCBI) via the Bio.Entrez and Bio.Blast.NCBIWWW modules to search and fetch biological records and publications from ncbi.nlm.nih.gov.
  • [COMMAND_EXECUTION]: The skill provides patterns for executing external bioinformatics command-line utilities—including Clustal Omega, MUSCLE, and BLAST+—using the Python subprocess module with explicit argument lists. The instructions include warnings to validate file paths and avoid interpolating unsanitized user input into shell commands.
  • [PROMPT_INJECTION]: The skill processes untrusted biological data formats (FASTA, GenBank, PDB, and NCBI XML), which are potential surfaces for indirect prompt injection. 1. Ingestion points: Files are parsed using Bio.SeqIO, Bio.PDB, and Bio.Entrez.Parser. 2. Boundary markers: The skill relies on standard format delimiters without custom boundary reinforcement. 3. Capability inventory: The skill possesses file write access, network capabilities via Entrez, and subprocess execution for external tools. 4. Sanitization: The documentation explicitly advises validating inputs for subprocess calls and recommends Biopython 1.87+ to mitigate CVE-2025-68463 when parsing untrusted XML files.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 08:09 PM
Security Audit — agent-trust-hub — biopython