NVIDIA-AI-IOTOfficial·9 Agent Skills Included

jetson-device-skills

Jetson device diagnostics, LLM serving, and video codec workflows

Diagnoses live NVIDIA Jetson devices, audits memory, and reclaims RAM with safe headless-mode changes. Serves, tunes, and benchmarks LLMs on vLLM, SGLang, llama.cpp, and Ollama with Jetson-correct containers and flags. Installs, verifies, and benchmarks NVIDIA Video Codec SDK and PyNvVideoCodec encode and decode pipelines. Eliminates guesswork about SKUs, JetPack versions, and GPU memory by reading live device data.

110|19

↗

npx skills add NVIDIA-AI-IOT/jetson-device-skills --all -g -y

All Skills in This Repository (9)

Pure Emerald Level Indicators

📦 In Repo

NVIDIA-AI-IOT

jetson-inference-mem-tune

Selects Jetson inference runtime and memory flags from an audit JSON for LLM/VLM serving workloads.

jetson-llm-benchmark

Benchmark Jetson LLM serving and return structured latency and throughput metrics.

jetson-speculative-decoding

Configure speculative decoding for Jetson vLLM servers with EAGLE-3 heads or draft models.

jetson-package

Select Jetson-compatible containers and Python package indexes for GPU-native installs.

jetson-print-device-info

Capture a read-only Jetson device summary with model, L4T, power mode, kernel, uptime, and OS version.

jetson-diagnostic

Capture a read-only Jetson health snapshot and memory summary.

jetson-headless-mode

Plan reversible Jetson headless-mode memory reclamation from audit data with dry runs and approval before applying changes.

jetson-memory-audit

Measure Jetson DRAM and NvMap usage on live NVIDIA Jetson devices.

jetson-llm-serve

Deploy OpenAI-compatible LLM or VLM serving endpoints on Jetson devices.

Official

Advanced

Frequently Asked Questions

FAQPage Schema

How to install jetson-device-skills?▼

Run `npx skills add NVIDIA-AI-IOT/jetson-device-skills --all -g -y` in your terminal to install all skills globally on your Jetson device.

How to free up memory on a Jetson device?▼

The jetson-memory-audit skill measures live DRAM and GPU memory usage, then jetson-headless-mode safely disables the desktop and background daemons to reclaim RAM with reversible commands.

How to serve an LLM on NVIDIA Jetson?▼

The jetson-llm-serve skill provides ready-made vLLM and SGLang launch recipes matched to your Orin or Thor device, and jetson-inference-mem-tune picks the right memory flags to avoid OOM errors.

Can it benchmark LLM and video performance on Jetson?▼

Yes. jetson-llm-benchmark measures TTFT, ITL, and throughput for vLLM, llama.cpp, and Ollama, while jetson-video-benchmark measures hardware encode and decode FPS with structured JSON output.

Does it work with Claude Code and Cursor?▼

Yes. All skills follow the standard SKILL.md format and install into the skill directories of Claude Code, Codex, Cursor, and NemoClaw/OpenClaw.