adme-property-predictor
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [SAFE]: The skill performs intended cheminformatics calculations using established libraries without suspicious side effects.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests SMILES strings and CSV data as seen in
scripts/main.py. Boundary markers are present via the use ofrdkit.Chem.MolFromSmiles, which validates input as a chemical structure. The capability inventory is restricted to file I/O and mathematical calculations, with no access to dangerous system calls or network resources. Sanitization is achieved through strict structural validation of the molecular notation. - [METADATA_POISONING]: The
adme-property-predictor_audit_result_v2.jsonfile contains safety assertions such as 'security: PASS'. These claims were identified as data to be evaluated and were not treated as authoritative conclusions by the analyzer.
Audit Metadata