scgpt
Pass
Audited by Gen Agent Trust Hub on Aug 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references external model weights and vocabulary files hosted by the Wang Lab at the University of Toronto via their GitHub repository (github.com/bowang-lab/scGPT).
- [COMMAND_EXECUTION]: Provides instructions for submitting remote compute jobs that execute shell commands (e.g.,
python3 embed.py) on HPC or cloud providers. - [DATA_EXFILTRATION]: Documents the process of uploading local AnnData (
.h5ad) files to remote compute environments for high-memory processing tasks. This is a functional requirement for the skill's stated purpose. - [SAFE]: Analysis of the skill instructions and metadata revealed no evidence of prompt injection, obfuscation, or unauthorized sensitive file access. The operations described are consistent with bioinformatics research workflows.
Audit Metadata