tooluniverse-precision-medicine-stratification
Pass
Audited by Gen Agent Trust Hub on Mar 15, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect prompt injection surface identified. Ingestion points: User-provided inputs for 'Disease/condition', 'Genomic data', and 'Clinical parameters' as defined in SKILL.md. Boundary markers: Absent. The skill instructions do not define specific delimiters or instructions to ignore potential commands embedded within the processed medical data. Capability inventory: The skill calls numerous external medical and genomic APIs (e.g., OpenTargets, FDA, PharmGKB) and performs a file-write operation to generate the final stratification report (SKILL.md, Phase 9). Sanitization: Absent. No input validation or sanitization mechanisms are described for the patient data processed.
- [SAFE]: The skill integrates with multiple verified and well-known technology and medical services including the NIH (PubMed, ClinVar), FDA, Ensembl, and PharmGKB. All data retrieval and analysis steps use established medical research tools. No unauthorized network activity, credential theft, or malicious code execution patterns were identified.
Audit Metadata