A workbench guide for NVIDIA GR00T, a robot foundation model, covering its files, training, testing, serving, and use.

Use it to download GR00T models, fine-tune them, evaluate results, run inference, serve the model, convert checkpoints, and check system status.

It brings the model’s different operations and checks into one documented command workflow. It also helps prevent deployment problems involving credentials, model access, checkpoints, and CUDA compatibility.

groot — Skill for Claude Code and Codex by nebius

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent

npx skills add nebius/nebius-physical-ai --skill groot

Clone the repo

git clone --depth 1 https://github.com/nebius/nebius-physical-ai

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/nebius/nebius-physical-ai/groot.svg)](https://agentmods.dev/skills/nebius/nebius-physical-ai/groot)

Your own site

<a href="https://agentmods.dev/skills/nebius/nebius-physical-ai/groot"><img src="https://agentmods.dev/badge/skills/nebius/nebius-physical-ai/groot.svg" alt="Measured on agentmods" height="20"></a>

Per session 39 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 1,575 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits

NVIDIA SkillSpector pass 7 Sept 2026

How audits are shown

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00039	$0.01575
Opus 5	$0.00019	$0.00788
Sonnet 5	$0.00008	$0.00315
Haiku 4.5	$0.00004	$0.00158

Measured today against content hash 42d6f2486312, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

groot scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/tools/groot/SKILL.md · 126 lines

How it starts

The opening of the file, as written. The whole thing — 126 lines — stays where its author put it; the contents beside it link to each section on GitHub.

GR00T

When To Use

Use this skill for NVIDIA GR00T robot foundation model workbench changes, especially NGC/Hugging Face model handling, embodiment tags, checkpoint conversion, serving, inference, and validation.

Procedure

Start with the current command surface:
```
npa workbench groot --help
```
Use download to stage model artifacts and credentials, finetune for training, eval for offline scoring, serve and infer for runtime calls, and convert when transforming checkpoints for downstream use. For declarative N1.7 training, use workflows/testing/groot-1-7-finetune.yaml; its gpu_count is a positive single-node world size that controls both the scheduler allocation and the real trainer.
Use status, system-info, and list for operational checks. Keep ensure-ingress, register-byovm, reload-env, and cleanup-partial scoped to setup and recovery flows.
Preserve credential redaction for NGC, Hugging Face, S3, and SSH values.
Before deploy provisions or updates anything, validate actual Hugging Face access to both the selected GR00T checkpoint and its runtime-fetched nvidia/Cosmos-Reason2-2B dependency. The operator's HF token and upstream permissions are the only local gate for gated weights; do not add a manual acceptance flag or a model-check bypass.

Three-Tier Contract

CLI: list, deploy, download, finetune, eval, serve, infer, convert, status, and system-info are the main user commands.
SDK/API: keep model, checkpoint, and storage path normalization in shared helpers so service and CLI routes do not diverge.
YAML: groot-1-7-finetune.yaml calls workbench.groot.finetune in the stage's own image. Keep model/source pins, checkpoint S3 URIs, run ID, and GPU count explicit; counts above one must use the upstream torchrun path.

Routing And Validation

GR00T does not require RT cores for the standard model paths.
Route throughput-heavy training/eval to H100/H200 unless a command or image specifically requires another target.
CUDA 13 alignment is vendor-paced on NVIDIA x86_64 CUDA 13 and is not a Nebius infrastructure blocker.

Read the full file on GitHub · 126 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

today Changed 42d6f2486312
8d ago First seen · 126 lines · 39 tokens per session scan A 09c499f38dbe

Subscribe to this mod's changes

groot is a skill published in the GitHub repository nebius/nebius-physical-ai (27 stars, last pushed today), licensed Apache-2.0. It adds 39 tokens to every session and 1,575 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Data and AI Hardware and IoT

spark-environment-setup

Set up a working ML training/inference environment on NVIDIA DGX Spark (GB10, aarch64, CUDA 13). Use when installing PyTorch/Unsloth/TRL/vLLM on DGX Spark, hitting libcudart or wheel-ABI errors on aarch64, or choosing between NGC containers and bare pip installs.

wshobson/agents · 76 tokens

spark-memory-thermal-ops

Manage unified memory and thermals during long-running ML jobs on NVIDIA DGX Spark. Use when planning memory headroom for a training run on GB10, when a job OOMs on unified memory, or when monitoring temperature and power during multi-hour training.

wshobson/agents · 59 tokens

spark-training-gotchas

Preflight and diagnose the ten known failure modes for ML training on NVIDIA DGX Spark. Use when a training run on DGX Spark fails to start, OOMs below the 128GB limit, slows down mid-run, or before any multi-hour training job on GB10.

wshobson/agents · 63 tokens

llama-cpp

Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.

davila7/claude-code-templates · 76 tokens

amc-run-rtsp-calibration

Calibrate a new dataset from live RTSP camera streams via the AutoMagicCalib REST API. Use when the user provides RTSP URLs or asks to calibrate live cameras; VIOS records clips, AMC ingests them, then runs calibration.

NVIDIA/skills · 59 tokens

amc-run-video-calibration

Calibrates pre-recorded cam.mp4 datasets through the AutoMagicCalib REST API. Use for user-supplied local MP4s; route live RTSP streams to amc-run-rtsp-calibration.

NVIDIA/skills · 57 tokens