sue-sandbox-admin

What problem does it solve?

This Skill eliminates the manual overhead and inconsistency of managing backend sandbox configurations for DeepResearch SUE ML/HPC scale-up experiments, which commonly leads to broken connection contracts, failed job submissions, and wasted debugging time when onboarding or updating HPC clusters, GPU servers, or cloud compute backends.

Core Features & Use Cases

• Full Backend Lifecycle Management: Register new sandboxes, update connection configs for existing backends, delete retired sandboxes, and audit the entire registry for consistency, replacing legacy standalone sandbox setup and RunPod config skills.
• Multi-Backend Support: Handles all supported backend classes including Slurm HPC (LUMI, Snellius), direct GPU servers, API cloud providers (RunPod), and local-only environments with class-specific validation and default path derivation.
• Use Case: When onboarding a new LUMI HPC cluster for SUE experiments, this skill walks you through collecting required account and path values, generating git-ignored secret configs, validating SSH and Slurm access, and handing off to sue-config to update your workspace runtime settings.

Quick Start

Use the sue-sandbox-admin skill to register a new LUMI HPC sandbox for your DeepResearch scale-up experiment workflow.