ai-drug-design-scientist
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill demonstrates professional standards for its domain, including risk disclaimers, quality verification checklists, and systematic decision frameworks. No malicious intent or suspicious patterns were identified.
- [COMMAND_EXECUTION]: Code examples provided in the references (e.g., active learning setup) are limited to data processing using legitimate scientific Python libraries like RDKit and BoTorch, without invoking unsafe system calls or arbitrary shell commands.
- [DATA_EXFILTRATION]: There are no indicators of data exfiltration; the skill does not access sensitive local files (e.g., SSH keys, environment variables) or attempt to transmit data to non-whitelisted external endpoints.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process molecular strings (SMILES) and protein structures (PDB/AlphaFold). While these external data formats represent a potential injection surface, the specialized nature of the processing and the lack of high-privilege capabilities tied to the results minimize the risk of exploitation.
Audit Metadata