pharmaclaw-ip-expansion-agent
Pass
Audited by Gen Agent Trust Hub on Mar 21, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill makes network requests to api.patentsview.org (USPTO) and pubchem.ncbi.nlm.nih.gov (National Institutes of Health) to retrieve patent and chemical information. These are well-known public research services.
- [COMMAND_EXECUTION]: The skill executes local Python scripts to perform molecular analysis using the RDKit library and to generate reports. It also references a model download from a reputable research organization for NLP tasks.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes strings (such as chemical names and therapeutic queries) provided by other agents.
- Ingestion points: Processes smiles, therapeutic, and query parameters from the input JSON in scripts/agent.py.
- Boundary markers: No specific delimiters are used to isolate untrusted input within the processing logic.
- Capability inventory: Includes the ability to perform network requests, write files (reports), and modify a local SQLite database.
- Sanitization: Input data is used for database queries and NLP extraction without explicit validation or sanitization.
Audit Metadata