
NVIDIA-verified agent skills for CUDA-X, RAG, and Physical AI
Get PyTorch, Unsloth, TRL, and vLLM running on DGX Spark.
Clone or download it from github.com, then drop the folder into ~/.claude/skills/ (Claude Code) or your agent's skills directory.
Share this product's name and rating in your README or on your website.
Updates may be delayed by image caching.
Setup guide for ML training and inference on NVIDIA DGX Spark (GB10, aarch64, CUDA 13), including libcudart and wheel errors and the NGC container versus bare-metal choice.
Part of the wshobson/agents plugin collection.
Used it? Write the first review.

NVIDIA-verified agent skills for CUDA-X, RAG, and Physical AI

166 scientific research skills that turn any agent into a co-scientist
Manage unified memory and heat during long training runs.
70+ agent skills for local-first clinical NLP: de-identifying text, clinical NER, FHIR export and terminology mapping.
NVIDIA's skill that converts one CT DICOM series into a Hounsfield-unit NIfTI volume, with affine evidence in a JSON result.
NVIDIA's skill for turning a word problem into a clean LP, MILP or QP: decisions, constraints and objective, before any code.