"Use Newton physics engine APIs for robotics simulation, solvers,
Scanned 9/8/2026
Install to Claude Code
npx -y skills add VectorSpaceLab/AREX-Skill --skill newton --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Newton?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vectorspacelab-newton)More formats (shields.io, HTML) on the badges page.
---
name: newton
description: "Use Newton physics engine APIs for robotics simulation, solvers,
contacts, asset import/export, sensors, viewers, examples, and repository
maintenance."
disable-model-invocation: true
metadata:
disco-role: operating
license: Apache 2.0
---
# Newton repo skill
Use this skill when a task involves Newton, `newton-physics`, the Python `newton` package, Warp-backed physics simulation, robotics/contact/deformable solver workflows, URDF/MJCF/USD import, Newton examples, or Newton repository maintenance.
## Start here
1. Confirm install/backend state with `scripts/check_newton_env.py`.
2. For a tiny public simulation check, run `scripts/newton_smoke.py --device cpu --steps 2`.
3. Route to the focused sub-skill below.
4. If the user is editing Newton source, also read `references/development-maintenance.md`.
5. If behavior depends on version-sensitive details, read `references/repo-provenance.md`.
## Installation and backends
Read `references/install-and-backends.md` before installing optional extras or diagnosing CUDA/RTX/Torch/USD/MuJoCo issues. Base Newton requires `warp-lang`; optional workflows use extras such as `newton[sim]`, `newton[importers]`, `newton[examples]`, `newton[onnx]`, `newton[rtx]`, `newton[torch-cu12]`, `newton[torch-cu13]`, and `newton[notebook]`.
Use CPU for first smoke checks unless the user specifically asks for CUDA, RTX, or Torch behavior. A CPU pass does not prove GPU performance.
## Sub-skill routing
| Task | Read |
| --- | --- |
| Build a model, add bodies/shapes/joints, allocate states/controls/contacts, replicate worlds, or write a minimal simulation loop | `sub-skills/modeling-simulation/SKILL.md` |
| Choose/configure solvers, CollisionPipeline, MuJoCo contacts, SDF/hydroelastic contacts, contact material tuning, or deterministic/GPU solver behavior | `sub-skills/solvers-contacts/SKILL.md` |
| Import URDF/MJCF/USD assets, diagnose optional importer deps, schema resolvers, mesh/remesh/heightfield utilities, or USD/File export artifacts | `sub-skills/asset-import-export/SKILL.md` |
| Use actuators, controllers, IK, `ArticulationView`, robot target layouts, sites for robotics control, or ONNX/Torch policy dependencies | `sub-skills/robotics-control/SKILL.md` |
| Use sensors, viewer backends, recording/replay, example CLI/browser, `--viewer`, `--device`, `--test`, headless visualization, or benchmark flags | `sub-skills/sensors-visualization/SKILL.md` |
| Diagnose cross-cutting install/import/backend or skill-staleness issues | `references/troubleshooting.md`, `references/repo-provenance.md` |
| Edit Newton source, tests, docs, examples, public API, or changelog | `references/development-maintenance.md` |
## Safe public commands
From this generated skill directory:
```bash
python scripts/check_newton_env.py --show-optional
python scripts/newton_smoke.py --device cpu --steps 2
python scripts/list_newton_examples.py --limit 20
```
Against an installed Newton environment, common package commands are:
```bash
python -m newton.examples --list
python -m newton.examples basic_pendulum --viewer null --device cpu --test
```
Use the user's Python environment for these commands. Do not assume this generated skill's construction environment is available.
## Operating rules
- Use public imports: `import newton`, `import warp as wp`, and public submodules such as `newton.solvers`, `newton.sensors`, `newton.viewer`, `newton.utils`, `newton.ik`, and `newton.actuators`.
- Do not import from `newton._src` in user-facing code, docs, examples, or generated answers.
- Set `newton.use_coord_layout_targets = True` before constructing new robotics models that write position targets.
- Treat MuJoCo, USD/importers, RTX, ONNX, Torch, notebooks, and viewer stacks as optional extras; check dependencies before running examples.
- Prefer tiny CPU smokes before long examples, GPU runs, downloads, or visualization.
- Never hide network, credential, external-service, or broad dependency installation behind a helper script.
## Staleness and provenance
`references/repo-provenance.md` records the source commit, package version, evidence paths, and verification summary used to create this skill. Refresh this skill when Newton's public APIs, docs, examples, dependency extras, or solver/importer behavior change.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!