Language:JapaneseEnglish

nvidia-skills Claude Code Plugin

Author
NVIDIA
Category
Development
Topics
AI Agents & AI App Development · Claude Code Customization & Workflow · Design & Media Production
First cataloged
2026-07-09 (UTC)
Explanation last updated
2026-09-03 (UTC)
Source (GitHub) last updated
2026-07-29 (UTC) (38 days ago)

The explanation below is AI-generated. Please verify it against the sources.

This plugin functions as a directory that helps a user locate the appropriate official NVIDIA skill from a large catalog of NVIDIA-authored skills. According to the homepage, the catalog lists 349 skills spanning domains such as AI and machine learning, Physical AI, infrastructure, accelerated computing, and developer tools, and can be filtered by domain, library, and target audience. Individual skills cover topics including RAG deployment, cuOpt optimization and routing, CUDA-Q, DeepStream video pipelines, Omniverse simulation and USD workflows, NeMo Retriever document search, DALI data loading, Nemotron model customization, cuDF GPU dataframes, and cuPyNumeric. Each skill can be installed separately into a Claude Code or Codex environment using a command-line install command shown per skill. The README states that this plugin bundle is generated from a configuration file for distribution to official marketplaces, rather than being part of NVIDIA's self-hosted marketplace.

About the service

NVIDIA maintains a broad set of official skills tied to its AI and GPU software stack, including products such as CUDA-Q, cuOpt, DeepStream, Omniverse, NeMo, Nemotron, and DALI. Some listed skills, such as those for AI-Q Blueprint, are described as requiring a reachable NVIDIA backend service to function, indicating that parts of this catalog connect to NVIDIA-hosted infrastructure rather than operating purely locally.

What you can do with nvidia-skills

  • Browse and filter the full catalog of official NVIDIA skills by domain, library, and audience
  • Install a specific matched skill into Claude Code or Codex using a shown command
  • Use skills for deploying, configuring, or troubleshooting RAG Blueprint services
  • Work with cuOpt for routing, optimization formulation, installation, or server/API usage
  • Access skills for CUDA-Q setup, DeepStream video pipeline development, and Omniverse/USD simulation tasks
  • Use skills for NeMo Retriever document search, DALI data loading, Nemotron customization, cuDF, and cuPyNumeric workflows

Sources

History of nvidia-skills

  • Claude Code Plugin Updated the official description of nvidia-skills

Back to list