pathml

Pass

Audited by Gen Agent Trust Hub on Jul 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADS
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the pathml library and its dependencies using standard package managers like uv pip. This is the intended primary library for the toolkit.
  • [DATA_EXFILTRATION]: The skill includes functionality for remote cell segmentation via the DeepCell API (deepcell.org). This is a documented feature for users who lack local GPU resources and targets a well-known academic service in the field of computational biology.
  • [COMMAND_EXECUTION]: The documentation provides examples of typical data management operations, including the use of Data Version Control (DVC) and batch processing scripts for HPC clusters (SLURM). These are standard practices for large-scale medical imaging research.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external medical imaging data (WSI) and clinical metadata. While this represents an ingestion surface for untrusted data, the capabilities described are limited to image analysis and data transformation without evidence of unsafe prompt interpolation or autonomous command execution triggered by data content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 12, 2026, 11:41 PM
Security Audit — agent-trust-hub — pathml