Skill profile
nvidia-skills
NVIDIA's own router across GPU acceleration, CUDA, agent training, inference, robotics, Physical AI, and simulation workflows.
What it actually does
- Acts as a router across NVIDIA's own catalog: GPU acceleration, CUDA, agent training, data loading, inference, robotics, Physical AI, Omniverse, and simulation workflows.
When to use it — and when not to
Use it when
- Your work touches GPU/CUDA programming, robotics, or NVIDIA's Physical AI stack.
Not for
- None of the above apply to your project — it's genuinely irrelevant otherwise.
Agent compatibility
- Claude Code
- Codex
NVIDIA's own repository states it is built to work in both Claude Code and Codex.
- Cursor— portable
Published to the open SKILL.md standard; not NVIDIA-declared Cursor support specifically.
Confirmed means the skill’s own source or vendor states it works with that agent. Portable means it only reaches that agent through the shared SKILL.md format and a directory convention — real, but a weaker claim. An agent with no entry here isn’t known to be unsupported; it simply isn’t documented either way, so nothing is claimed. Full model on Agent Skills Compatibility.
Evidence and provenance
Maintained by NVIDIA (vendor). Primary source: NVIDIA/skills, last checked against this page’s claims on . This page summarizes and analyzes that source — it isn’t a copy of its SKILL.md or README, and installation happens at the source, not here.
Also covered in The Best Claude Code Skills, part of the Claude Code guide.