run-experiment
Pass
Audited by Gen Agent Trust Hub on May 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill generates shell scripts from local templates and executes them locally or on remote servers via SSH to launch user-defined machine learning experiments.
- [DATA_EXFILTRATION]: Uses scp to transfer generated job scripts to remote login nodes on HPC clusters as part of the intended job submission workflow.
Audit Metadata