cibersort-immune-infiltration-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: Comprehensive analysis across all threat categories (including prompt injection, data exfiltration, and remote code execution) found no malicious intent or security vulnerabilities. The code is well-structured, follows least-privilege principles, and includes robust error handling and logging.
- [EXTERNAL_DOWNLOADS]: The documentation references standard scientific R packages from CRAN and Bioconductor (such as preprocessCore and e1071). These are well-known, established repositories for statistical software and are treated as safe sources.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided CSV/TSV files containing expression matrices and group labels. It parses these files numerically and applies strict validation for finite values and numeric types, minimizing the risk of data-driven prompt injection or schema confusion.
- [COMMAND_EXECUTION]: The skill operates via local R scripts. Command execution is limited to the test suite using standard R functions to run its own entry point for validation purposes, with no evidence of arbitrary or dangerous system calls.
Audit Metadata