NVIDIA Jetson Device Skills

NVIDIA Jetson Device Skills is a catalog of agent skills for working with a live NVIDIA Jetson device after it has booted. You can also discover these skills through the NVIDIA Agent Skills catalog, alongside skills for other NVIDIA products.

The skills provide agent-readable instructions and helper scripts for diagnostics, memory optimization, package selection, LLM serving, and inference and video benchmarking. Use them with Claude Code, Codex, Cursor, or other compatible coding agents to carry out workflows grounded in your live device’s configuration.

For host-side BSP customization and flashing workflows, see Jetson BSP Skills.

Start Here

New to AI-assisted development on Jetson?

• Hands-on walkthrough: AI-assisted development on Jetson — Follow a guided workflow from device preparation to a live VLM demo using a coding agent.
• Jetson Agent Skills guide — Learn how skills work, choose between Device and BSP Skills, and explore installation options.
• NVIDIA Agent Skills catalog — Discover Jetson skills alongside skills for other NVIDIA products.

This project is currently not accepting contributions.

Installation

Clone the repository on the Jetson device:

git clone https://github.com/NVIDIA-AI-IOT/jetson-device-skills.git
cd jetson-device-skills

Install the skills into the agent skill directories on the Jetson:

./install.sh

By default, install.sh links the skills into the user-level skill roots for Claude Code, Codex, and Cursor:

• ~/.claude/skills
• ~/.codex/skills
• ~/.agents/skills
• ~/.cursor/skills

You can select specific targets or copy files instead of creating symbolic links:

./install.sh --targets claude,cursor
./install.sh --targets cursor-project --project /path/to/project
./install.sh --copy
./install.sh --force

For NemoClaw/OpenClaw sandboxes:

./install.sh --targets nemoclaw --nemoclaw-sandbox jetson-skills

Restart the agent session after installation so the new skill entries are picked up. If your agent expects a different skill directory, copy or sync the skills/ directory into that location. Keep each skill as a complete directory containing its SKILL.md, scripts/, and references/ content.

Included Skills

Skill	What it helps you do
jetson-diagnostic	Inspect device identity, memory, GPU activity, thermals, power, storage, and services.
jetson-memory-audit	Measure memory usage and verify the effect of memory reclamation.
jetson-headless-mode	Configure a headless Jetson by managing desktop and background services.
jetson-inference-mem-tune	Choose inference runtimes and memory settings for the device.
jetson-llm-serve	Configure Jetson-appropriate LLM serving workflows.
jetson-llm-benchmark	Measure LLM inference performance with structured benchmark results.
jetson-package	Select Jetson-compatible packages, wheels, and containers.
jetson-speculative-decoding	Explore Jetson-specific speculative decoding configurations.
jetson-video-setup	Install and verify native Video Codec SDK and PyNvVideoCodec components.
jetson-video-capability	Check live encode/decode capabilities against documented device support.
jetson-video-recipe	Translate video codec requirements into validated configuration plans.
jetson-video-benchmark	Measure encode and decode performance on the device.
jetson-video-pipeline	Execute and verify video codec pipeline stages.

Usage

Each skill lives under skills/<skill-name>/ and starts with a SKILL.md file. After install.sh links or copies the skills into the agent's skill directory, agent runtimes such as Cursor, Claude Code, Codex, or NemoClaw/OpenClaw can discover the skills from their frontmatter descriptions and follow the instructions in the selected skill.

Some skills include helper scripts under scripts/, but users normally do not need to call those scripts directly. The agent should invoke the relevant helper when a skill asks for live Jetson data, then use the returned output as the source of truth.

Repository Layout

jetson-device-skills/
├── README.md
├── LICENSE
├── install.sh
├── agents/
└── skills/
    ├── jetson-diagnostic/
    ├── jetson-headless-mode/
    ├── jetson-inference-mem-tune/
    ├── jetson-llm-benchmark/
    ├── jetson-llm-serve/
    ├── jetson-memory-audit/
    ├── jetson-package/
    ├── jetson-speculative-decoding/
    ├── jetson-video-benchmark/
    ├── jetson-video-capability/
    ├── jetson-video-pipeline/
    ├── jetson-video-recipe/
    └── jetson-video-setup/

Related Resources

• Jetson AI Lab — Models, tutorials, and applications for generative AI on Jetson.
• NVIDIA Agent Skills — The central NVIDIA skills catalog.
• Jetson BSP Skills — Skills for host-side BSP customization and flashing workflows.
• Report an issue — Report a bug or documentation problem. For security issues, follow SECURITY.md.

License

This code is dual-licensed with documentation under the CC-BY-4.0 AND source code under Apache-2.0 license terms. See LICENSE.