tooluniverse-protein-therapeutic-design
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill explicitly directs the agent to 'write and run Python code via Bash' for statistical analysis, data processing, and scoring. This is an intended and legitimate use of the agent's computational tools to ensure scientific accuracy.
- [EXTERNAL_DOWNLOADS]: The skill fetches protein structures from the Research Collaboratory for Structural Bioinformatics (RCSB) at
files.rcsb.org. This is a well-known, trusted, and necessary service for obtaining structural data for protein design. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from external scientific databases (PDB, UniProt, EMDB). While this presents a theoretical surface for indirect prompt injection, these are curated repositories and the skill uses the data as input for standard bioinformatics tools without demonstrated risk of instruction override.
- Ingestion points: Structural and sequence data retrieved via
UniProt,PDBe, andEMDBtools. - Boundary markers: None explicitly defined for external biological data.
- Capability inventory: Bash/Python execution for scientific computation; file creation for design reports and FASTA sequences.
- Sanitization: Data is processed through established bioinformatic schemas; no explicit NL sanitization is required for these data types.
Audit Metadata