What problem does it solve?
This Skill orchestrates the multi-phase cluster onboarding process by gathering SSH details, writing .raca/clusters.yaml entries, verifying connectivity, and checking GPU access so engineers never miss a required configuration step.
Core Features & Use Cases
- Cluster type triage: Ask whether the target is SLURM, RunPod, or a local GPU and route to the appropriate workflow so onboarding never accidentally uses the wrong backend.
- SLURM deep validation: Collect host, user, VPN/2FA requirements, discover partitions/accounts via sacctmgr and sbatch --test-only, record scratch, modules, and slurm prefix details, then test with raca auth/ssh.
- RunPod and local alternatives: Validate the RunPod API key with curl, capture GPU details via nvidia-smi, and confirm local CUDA readiness so any compute source is production-ready.
Quick Start
Ask the assistant to configure a new SLURM, RunPod, or local GPU cluster by providing the nickname, hostname, user, and necessary authentication details.