tooluniverse-polygenic-risk-score
Pass
Audited by Gen Agent Trust Hub on Mar 8, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious code, obfuscation, or prompt injection patterns were detected across the skill's files. The behavior of the skill matches its stated purpose.
- [DATA_EXFILTRATION]: The skill handles sensitive genetic information; however, this data is processed locally within the agent's environment using mathematical formulas and is not transmitted to unauthorized external domains. Network activity is limited to fetching public GWAS data via the tooluniverse library.
- [EXTERNAL_DOWNLOADS]: Dependencies are restricted to the tooluniverse framework and the pyvcf library, both of which are legitimate tools for bioinformatics and are used appropriately.
- [PROMPT_INJECTION]: The skill has an ingestion surface for external data (genotypes and GWAS associations). The risk of indirect prompt injection is minimal as data is parsed into structured numeric formats (e.g., beta coefficients, p-values) and used in calculations rather than being passed as free text for execution.
Audit Metadata