tooluniverse-clinical-trial-design
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches clinical trial metadata and regulatory history from official government and scientific repositories, including clinicaltrials.gov and api.fda.gov. These interactions target well-known and trusted services.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external scientific databases and summarizes it into markdown reports, which represents an indirect prompt injection surface if external records contain adversarial content.
- Ingestion points: Tool outputs from PubMed, DrugBank, and OpenTargets processed in python_implementation.py and trial_pipeline.py.
- Boundary markers: Absent; tool responses are directly summarized into report files.
- Capability inventory: Outbound network GET requests via the requests library and local file writing via the open function in python_implementation.py. Note that output filenames are constructed from user-supplied drug names without path sanitization.
- Sanitization: Absent; external research data is not escaped or validated before being written to generated reports.
Audit Metadata