hyperpod-slurm-debugger

Pass

Audited by Gen Agent Trust Hub on Jun 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The script executes local aws sagemaker CLI commands to retrieve cluster metadata and node lists. These commands use validated cluster names and regions.
  • [REMOTE_CODE_EXECUTION]: The skill uses aws ssm start-session to execute diagnostic commands on remote cluster nodes. Remote payloads are constructed using environment variable exports that are safely quoted using jq @sh to prevent command injection, and the entire payload is base64-encoded for secure transport via the SSM agent.
  • [DATA_EXPOSURE]: The diagnostic script reads Slurm configuration files (slurm.conf, slurmdbd.conf) and HyperPod metadata (provisioning_parameters.json) to identify configuration desyncs. This access is scoped to service-specific diagnostic data within the cluster environment.
  • [PROMPT_INJECTION]: The instructions explicitly constrain the agent to a "read-only" and "diagnostic-only" mode, prohibiting the recommendation or execution of state-mutating commands. It directs users to official AWS and Slurm documentation for remediation.
  • [EXTERNAL_DOWNLOADS]: The skill references documentation and troubleshooting guides from official AWS and Slurm (SchedMD) domains, as well as the official HyperPod cluster setup repository on GitHub.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 15, 2026, 01:10 AM
Security Audit — agent-trust-hub — hyperpod-slurm-debugger