tao-run-on-slurm

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides detailed instructions for executing shell commands on remote systems using SSH. This includes submitting jobs via sbatch, monitoring with squeue and sacct, and managing jobs with scancel. These operations are fundamental to its purpose of cluster job management and are documented with security considerations such as non-interactive mode.
  • [CREDENTIALS_UNSAFE]: The skill handles sensitive credentials including NGC_KEY, HF_TOKEN, and SSH private keys. It incorporates explicit security mitigations, such as using printf heredocs to ensure secrets are not stored in shell history or displayed in output, and enforcing restricted file permissions (chmod 600) for identity files and credential storage.
  • [DYNAMIC_EXECUTION]: The skill generates job specification files and .sbatch scripts from templates at runtime. To mitigate risks, it includes steps to validate these generated scripts using a secret-redacting linter (redact_secrets.py) and shell syntax checking (bash -n) before they are submitted for execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 02:56 AM
Security Audit — agent-trust-hub — tao-run-on-slurm