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.
npx agentmods add skills/robium-ai/robium/2.1.1npx skills add robium-ai/robium --skill 2.1.1git clone --depth 1 https://github.com/robium-ai/robiumWrote 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.
[](https://agentmods.dev/skills/robium-ai/robium/2.1.1)<a href="https://agentmods.dev/skills/robium-ai/robium/2.1.1"><img src="https://agentmods.dev/badge/skills/robium-ai/robium/2.1.1.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00162 | $0.06644 |
| Opus 5 | $0.00081 | $0.03322 |
| Sonnet 5 | $0.00032 | $0.01329 |
| Haiku 4.5 | $0.00016 | $0.00664 |
Grade B, and why
lerobot scanned grade B with 1 finding 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 5d ago.
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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
`ffmpeg` for video decoding — `sudo apt install ffmpeg` (Linux) or How it starts
The opening of the file, as written. The whole thing — 396 lines — stays where its author put it; the contents beside it link to each section on GitHub.
lerobot
The manipulation-vertical core tool skill for robium: the LeRobotDataset
format, loading and recording episodes, training imitation-learning and
VLA policies (ACT, Diffusion, Pi0/Pi0.5/SmolVLA and others), evaluating in
simulation, and teleoperation. LeRobot (huggingface/lerobot, PyPI package
lerobot, version 0.6.0 as of 2026-07-12 (PyPI-verified; re-check —
this number goes stale fast), requires-python >=3.12)
is HuggingFace's end-to-end robot-learning library — this skill embeds the
robotics-specific glue (dataset shape, training/eval CLI, sim envs,
teleoperation) and delegates hub mechanics (auth, upload/download, model
cards) to the huggingface skill's territory. LeRobot moves fast; every
command below is either a direct upstream docstring/doc example fetched
on 2026-07-10 or marked with how it was verified — re-check before relying
on exact flags in a real project.
When to use this skill
- Any manipulation/imitation-learning task: loading or recording a LeRobotDataset, training a policy (ACT, Diffusion, Pi0-family, SmolVLA, ...), evaluating a policy in simulation, teleoperating a robot arm.
- The trigger phrases in the description: 'lerobot', 'manipulation policy', 'imitation learning', 'train a robot arm policy', 'ACT', 'diffusion policy'.
- Cross-references — go to the sibling skill instead when the question is:
- Hub auth, dataset/model upload-download, model cards, repo management →
the
huggingfaceskill's territory; LeRobot's ownhf auth login/hf uploadcommands are shown here only where a lerobot workflow requires them inline. - Whether to use uv or Docker, GPU passthrough, headless/remote display →
environments(load first if not already decided; see Key directives). - Deciding which dataset(s) to source or combine for a task → the
dataumbrella skill's territory. - Rendering/inspecting recorded episodes in depth → the
rerunskill; LeRobot's own--vizextra andlerobot-dataset-vizscript wrap Rerun directly — cross-referenced here by name, not re-taught. - Classical motion planning (no learned policy) → out of scope repo-wide; this skill and the manipulation vertical are learning-based only.
- The NVIDIA Isaac Lab RL stack (GPU-parallel RL/IL training environments,
prebuilt tasks, policy export) →
isaac-lab. - The whole-stack decision this feeds into →
architect(routes here).
- Hub auth, dataset/model upload-download, model cards, repo management →
the
What ships with it
6 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 5d ago First seen · 396 lines · 162 tokens per session scan B 64741341d2b9
lerobot is a skill published in the GitHub repository robium-ai/robium (9 stars, last pushed 7d ago), licensed MIT. It adds 162 tokens to every session and 6,644 once invoked, about $0.0008 per session on Opus 5. A static security scan graded it B with 1 finding (asks for root). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
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.
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.
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.
local-inference-tuning
Select and tune a local LLM inference engine for the user's hardware. Use when setting up or auditing local/private model serving, choosing between MLX, llama.cpp, Ollama, and vLLM, estimating model fit, deciding cache/storage policy, tuning batching and KV cache flags, running smoke benchmarks, or exposing an…
yolo-export
Use when exporting or deploying Ultralytics YOLO models in Platform or code — the Platform Export tab and yolo export/model.export() for ONNX, TensorRT, CoreML, Core AI, OpenVINO, LiteRT, NCNN, ExecuTorch, and NPUs (RKNN, QNN, Hailo, Ascend, IMX, Axelera, DeepX), FP16/INT8 quantization, benchmarking, and non-Python…
gpu-debian-setup
Install and verify NVIDIA proprietary GPU drivers on Debian (including trixie/13) so local LLM tools (Ollama, llama.cpp, vLLM) can use the GPU. Covers nouveau blacklist, initramfs, the trixie nvidia-smi transitional-dummy gotcha and its official .run extraction fix, and verifying Ollama actually uses VRAM. Use…