iupac-name-identification-biot5
Pass
Audited by Gen Agent Trust Hub on Mar 25, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a straightforward workflow for chemical nomenclature using the open_biomed framework. All operations are confined to chemical data processing and model inference.
- [COMMAND_EXECUTION]: No unauthorized or dangerous shell commands are executed. The example script uses sys.path manipulation to locate the local library, which is a standard development practice for package-internal examples.
- [DATA_EXFILTRATION]: No sensitive data exposure or exfiltration patterns were detected. The skill interacts with PubChem via standard tools to retrieve molecule data based on common names.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided SMILES strings, molecule names, and natural language questions. While this constitutes an ingestion surface for external data, the resulting model outputs are treated as literal text (IUPAC names) and are not used to drive sensitive system capabilities or code execution.
- Ingestion points: SMILES strings, molecule names, and custom questions provided via CLI arguments in basic_example.py.
- Boundary markers: None implemented in the example script.
- Capability inventory: Limited to text output and chemical data retrieval.
- Sanitization: None observed; the skill relies on the underlying model's inherent safety filters and the restricted nature of the output domain (chemical nomenclature).
Audit Metadata