pathml
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Installs the legitimate 'pathml' toolkit and its dependencies from PyPI. It also utilizes the DeepCell API for remote cell segmentation tasks, which is an established service in the bioinformatics field.
- [COMMAND_EXECUTION]: Provides examples for generating and submitting SLURM job scripts to HPC clusters, which is a standard procedure for large-scale pathology workflows.
- [DATA_EXFILTRATION]: Features a SegmentMIFRemote transform that sends image data to the DeepCell API for processing, which is a legitimate and documented functionality for remote segmentation.
- [SAFE]: Indirect Prompt Injection Surface Analysis: Ingestion points: slide loading in references/image_loading.md and clinical CSV reading in references/data_management.md. Boundary markers: none. Capability inventory: file writing, shell execution (SLURM), and network requests. Sanitization: none. The skill is a technical toolkit and exhibits no patterns of malicious instruction following from untrusted data.
- [SAFE]: No obfuscation, malicious instructions, or credential exposure were detected. The skill correctly represents the PathML library's intended scientific use cases.
Audit Metadata