tooluniverse-clinical-trial-design
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill retrieves data from well-known biomedical and regulatory services including OpenTargets, ClinVar, gnomAD, PubMed, DrugBank, the FDA, and ClinicalTrials.gov to populate feasibility reports.
- [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill ingests untrusted data from external literature abstracts (PubMed) and clinical trial records (ClinicalTrials.gov) based on user-supplied indication and drug names as seen in the search logic of
python_implementation.py. - Boundary markers: No explicit delimiters or instructions are present to separate retrieved external content from the agent's primary instructions during report synthesis.
- Capability inventory: The skill includes scripts capable of performing network requests for data retrieval and writing synthesized results to local markdown files.
- Sanitization: User input is interpolated directly into tool query parameters without explicit sanitization or escaping logic.
- [COMMAND_EXECUTION]: The skill provides pre-defined Python scripts (
python_implementation.pyandtrial_pipeline.py) that are intended to be executed to automate the research workflow and generate feasibility reports.
Audit Metadata