NVIDIANVIDIAOfficialยท150 Agent Skills Included

skills

Official NVIDIA skills for GPU, AI, and accelerated computing workflows

Guides AI agents through NVIDIA software stacks including CUDA-Q, cuOpt, cuDF, DOCA, DeepStream, TAO, NeMo, and Jetson. Eliminates guesswork when installing SDKs, training models, deploying GPU pipelines, or debugging BlueField and Jetson hardware. Each skill ships verified instructions, scripts, and benchmarks so agents complete NVIDIA workflows correctly on the first attempt.
npx skills add NVIDIA/skills --all -g -y
Available:

Tells the AI agent the layout rules and output constraints for the DICOM metadata extraction skill, including schema alignment and which values it may not infer.

All Skills in This Repository (150)

Pure Emerald Level Indicators
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

nemotron-voice-agent-deploy

Deploy Nemotron Voice Agent on Workstation, Jetson Thor, or Cloud NIMs with WebRTC or WebSocket transport.

Official
Advanced
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

ptq

Quantize pretrained models to nvfp4, fp8, or int4_awq checkpoints using ModelOpt PTQ.

Official
Advanced
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

deployment

Deploy HuggingFace model checkpoints as OpenAI-compatible endpoints with vLLM or SGLang.

Official
Advanced
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

evaluation

Generate and run LLM evaluation configurations with NeMo Evaluator Launcher.

Official
Advanced
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

debug

Execute commands inside a remote Docker container via a file-based relay.

Official
Intermediate
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

add-benchmark

Scaffold NeMo-Gym benchmarks with resources servers, data pipelines, and YAML configs.

Official
Advanced
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

cudaq-guide

Guide users through installing, configuring, and using CUDA-Q in Python and C++.

Official
Intermediate
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

byob

Automate end-to-end creation and evaluation of custom LLM benchmarks with BYOB.

Official
Advanced
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

cuopt-qp-api-cli

Solve quadratic programming problems from the command line using cuOpt's CLI.

Official
Intermediate
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

cuopt-lp-milp-api-python

Formulate and solve LP and MILP problems with cuOpt's Python API.

Official
Intermediate
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

cuopt-installation-common

Identify cuOpt system requirements for GPU, CUDA, and deployment environments.

Official
Intermediate
๐Ÿ“ฆ In Repo
NVIDIANVIDIA

cuopt-lp-milp-api-cli

Solve LP and MILP problems from MPS inputs via the cuOpt CLI.

Official
Intermediate

Frequently Asked Questions

FAQPage Schema
How to install NVIDIA skills?โ–ผ

Run `npx skills add NVIDIA/skills --all -g -y` in your terminal to install the full NVIDIA skill catalog globally.

What do NVIDIA agent skills do?โ–ผ

They teach AI coding agents how to correctly use NVIDIA products like CUDA-Q, cuOpt, DOCA, DeepStream, TAO, NeMo, and Jetson, covering setup, training, deployment, and debugging.

Which AI agents work with NVIDIA skills?โ–ผ

All skills follow the universal SKILL.md standard and run in Claude Code, Codex, Cursor, Kiro, and other compatible agents.

How do I find the right NVIDIA skill for my task?โ–ผ

Install the nvidia-skill-finder skill, which detects NVIDIA-related requests and recommends the matching skill from the live catalog before you install anything.

Are NVIDIA skills verified before release?โ–ผ

Yes. Every skill passes a three-tier evaluation covering security, correctness, discoverability, and effectiveness, with results published in each skill's BENCHMARK.md file.

Related Repositories in Software Engineering

View All in Software Engineeringโ†’