tooluniverse-precision-medicine-stratification
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates within the expected parameters of a medical and bioinformatics analysis tool. All operations described in the documentation and implemented in the test scripts align with the primary purpose of precision medicine stratification.
- [EXTERNAL_DOWNLOADS]: The skill integrates with over 50 reputable external biological and clinical databases, including OpenTargets, NCBI (ClinVar, PubMed), Ensembl, PharmGKB, and FDA Open Data. These connections are essential for retrieving genetic associations, drug indications, and clinical guidelines.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes untrusted user-provided genomic variants and clinical data.
- Ingestion points: Phase 1 involves parsing free-text disease names and variant information from user prompts.
- Boundary markers: The workflow utilizes structured phases and internal logic to categorize and resolve inputs before processing.
- Capability inventory: The skill executes network-based tool calls for data retrieval and performs local file writing for report generation.
- Sanitization: Input data is validated and normalized against established ontologies (EFO/MONDO) and official gene symbols (Ensembl/HGNC) in Phase 1, which serves as a protective layer against malicious payloads.
Audit Metadata