bio-proteomics-differential-abundance
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied proteomics intensity matrices and experimental metadata, which introduces a surface for indirect prompt injection findings.
- Ingestion points: External data enters the agent context via the
protein_matrixandsample_infoinputs in R code and theintensitiesDataFrame in Python examples. - Boundary markers: No specific delimiters or instructions are provided to the agent to treat input data as untrusted or to ignore instructions embedded within values.
- Capability inventory: The skill's capabilities are limited to statistical calculations; it lacks any tools for network communication, file system writing, or arbitrary shell command execution.
- Sanitization: While the data is transformed through log-scaling and filtering, there is no security-specific sanitization of string-based metadata or headers against potential injection attacks.
- [EXTERNAL_DOWNLOADS]: The documentation provides standard installation instructions for well-known bioinformatics libraries from public registries such as PyPI and Bioconductor. These references are routine for the domain and target reputable services.
Audit Metadata