duke-cs-cluster-skill

Check current availability of GPUs on the Duke CS cluster. Use when the user asks about cluster load, GPU contention, wait times, or wants to know if a specific GPU type is free before submitting a job. Optional argument is a GPU type (a5000, a6000, rtx_pro_6000, v100, p100, rtx_2080, rtx_5000).

Trigger before submitting any Slurm job (sbatch/srun), when diagnosing why a job failed to schedule, or when choosing GPU types, partitions, storage paths, or resource limits on the Duke CS compute cluster. Do NOT trigger for non-Slurm tasks or clusters other than Duke CS.

Show the current user's running and pending Slurm jobs on the Duke CS cluster, with resource totals and warnings if approaching per-user caps. Use when the user asks about their own jobs, what's running, or how close they are to their GPU/CPU/job limits.