pharmaclaw-catalyst-design

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill content and metadata are consistent with its stated utility in medicinal chemistry research.
  • No prompt injection or behavior override patterns were found in the instructions.
  • No evidence of obfuscation or hidden content was detected.
  • [SAFE]: The skill operates entirely on local resources.
  • Functions are implemented through local Python scripts: scripts/catalyst_recommend.py, scripts/ligand_designer.py, and scripts/chain_entry.py.
  • Data is retrieved from a local reference file: references/catalyst_database.json.
  • No network operations or external downloads from untrusted sources are specified in the documentation.
  • [SAFE]: Structured input handling.
  • The skill consumes JSON-formatted inputs for reaction types and substrate SMILES strings.
  • While this establishes a data ingestion surface, the absence of high-risk capabilities (such as unrestricted shell execution or outbound network calls) ensures a safe operating environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 04:43 AM
Security Audit — agent-trust-hub — pharmaclaw-catalyst-design