NVIDIA Agent Skills
The official catalog of NVIDIA agent skills: CUDA, accelerated libraries, RAG, robotics and simulation through coding agents
- 3,944 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Skills lead the agent to run GPU code, install software and deploy containers
- Some skills configure networking, DPUs and cluster deployments
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/nvidia-agent-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Skills lead the agent to run GPU code, install software and deploy containers
- Some skills configure networking, DPUs and cluster deployments
This is third-party code. Review the repository files before installing.
What it does
The repository is NVIDIA's official catalog of skills for coding agents. Skills are portable instruction sets that teach an agent to use NVIDIA software correctly: accelerated computing and data work (cuDF, cuPyNumeric, DALI), optimization (cuOpt), quantum computing (CUDA-Q), RAG and blueprints, deployment and inference (Dynamo), networking and DPUs (DOCA), robotics, simulation and Physical AI. Skills are maintained in the product repositories and mirrored here daily by an automated pipeline, so the list keeps growing. The set includes a skill finder that helps the agent pick the right NVIDIA skill for a task. The catalog installs through the skills CLI or as a plugin, and a skill is picked up when the agent meets a relevant task.
Who it is for. For developers, data and robotics engineers who work with the NVIDIA stack through a coding agent.
Good fit when
- You need to speed up data processing or compute on GPU via cuDF, cuPyNumeric or DALI
- You need to solve an optimization problem in cuOpt or build a RAG workflow on NVIDIA blueprints
- You need the agent's help with deployment, inference, robotics or simulation on NVIDIA software
Not a fit when
- You have no access to the required NVIDIA hardware, containers and services
- The task is not related to the NVIDIA stack
Example request
Solve a linear programming problem with cuOpt and show the finished Python codeLimitations
Skills are instructions; the actual work needs NVIDIA software, hardware and services: suitable GPUs, containers and images from NGC, and sometimes accounts and specialized hardware such as BlueField DPUs. Each skill has its own requirements for library and driver versions. Skills install from GitHub and via npm, but availability of some NVIDIA services and downloads from Russia without a VPN is not guaranteed. Apache 2.0 for code and CC-BY-4.0 for documentation.
How to disable. Remove installed skills via npx skills update and confirming removal, or delete their folders from the agent's skills directory. If installed as a plugin, remove it via /plugin.
FAQ
Are these official NVIDIA skills?
Yes, the repository is stated as an official, NVIDIA-verified catalog. Skills are maintained in product repositories and mirrored into this catalog daily.
Are the skills alone enough to do the work?
No. Skills give the agent methodology, but real tasks need GPUs, NGC containers and the NVIDIA software each skill targets.
Discussion
No questions or reviews yet.
Related
SkillsA set of independent skills against drift in large codebases: hard-cut, root-cause tracing, lean testing and CI right-sizing.
A library of 30-plus instruction skills for coding agents: planning, building, testing, hallucination guards and shipping
Tech Leads Club Agent Skills
Agent Skills
A vetted skill registry with a CLI that installs into twenty-plus agents and an MCP server for on-demand skill lookup
OpenClaw Agent Skills
openclaw-agent-skills
Shared skills for coding agents that work on OpenClaw projects: review and other workflows. A narrow topic for people who write OpenClaw code
Softaworks Agent Toolkit
Softaworks Agent Skills
A marketplace of skills, subagents and commands for coding agents: development, documentation, planning and workplace communication
A set of game-dev skills for AI agents with a router that picks the right ones for the engine and task
README in short
The README presents NVIDIA's official catalog of agent skills for Claude Code, Codex and other coding agents. Skills teach the agent to use NVIDIA software: accelerated computing, optimization, RAG, deployment, networking, robotics and simulation, and they are maintained in product repositories and mirrored here daily. Installation goes through the skills CLI with a single npx skills add nvidia/skills command; you can install individual skills, target a specific agent, and update or browse the catalog. A skill finder helps pick the right skill. Licensed under Apache 2.0 and CC-BY-4.0.
README badge
Original SKILL.md
---
name: accelerated-computing-cudf
description: Official NVIDIA-authored guidance for NVIDIA cuDF GPU DataFrames, pandas acceleration, dask-cuDF, ETL, joins, groupby, CSV/Parquet I/O, nullable semantics, and multi-GPU DataFrame workloads.
license: CC-BY-4.0 AND Apache-2.0
metadata:
author: NVIDIA
tags:
- cudf
- dataframes
- pandas
- dask-cudf
- etl
---
# cuDF & dask-cuDF Implementer's Guide
## Compatibility
- Requires NVIDIA Volta or newer on CUDA 12, or Turing or newer on CUDA 13. cuDF sweet spot: more than 100K rows.
## Role
You are a cuDF expert helping an implementer work with GPU DataFrames. Choose the path from the user's intent: cudf.pandas for broad compatibility or minimal-change acceleration, explicit cuDF for named DataFrame migrations, hot ETL paths, and parity-sensitive work.
## Critical Rules
1. Choose the right cuDF path.
2. Size gate: 100K rows minimum.
3. Keep conversions at boundaries.
4. Float32 is your friend.
5. Validate semantics on representative slices.