What does this skill do, and when should you use it?
This skill coordinates an end-to-end workflow from CAD or another source asset to a simulation-ready USD or USDZ package. It covers dependency preflight, source conversion, minimum viability checks, material and physics assignment, SimReady conformance, validation, rendering, and optional packaging. When property assignment is enabled, Content Agents services must be verified or deployed first. It is a documentation-driven orchestrator and does not provide one monolithic runner command.
Confirms the source path, resolves workflow intent and output location, runs preflight, and creates an environment manifest. It invokes installed stage references for conversion, USD validation, Content Agents, SimReady conformance, rendering, and packaging. It converts CAD or other source assets to USD or accepts existing USD input; assigns materials, physics, and optional textures; runs minimum, geometry, physics, and SimReady validation; repeats conformance when structured repairable requirements are reported; and emits final USD artifacts plus optional previews, package outputs, and consolidated reports.
- An engineering team needs to convert CAD parts into simulation assets with materials, physics, and SimReady validation.
- A user wants to smoke-test source conversion and validate the resulting USD without deploying Content Agents.
- A simulation developer has an existing USD asset and needs material or physics assignment followed by conformance checks.
- An asset team needs preview images and a validated deliverable package for publishing a final simulation asset.
How do you install this skill?
- Execution may process or upload user CAD assets through external Content Agents, converters, and rendering services; confirm data flow, credential scope, network reachability, and organizational compliance before use.
- Broad end-to-end requests default property_assignment_intent to run and require service deployment or verification; obtain explicit confirmation before deployment, external calls, or overwriting an existing deliverable.
- The benchmark reports an overall FAIL, Tier 3 is unavailable, and it identifies missing Examples, an overly long description, and multiple duplication findings; it is not evidence of successful execution or validated quality.
- Add reproducible tests, concrete maintenance ownership and changelog data, reconcile license metadata, and document mainland-China network and Chinese-language limitations.
- Shell / CLI
- Network access
- Local filesystem
Python 3.12uvDockerNVIDIA Container ToolkitNVIDIA GPUNVIDIA_API_KEY when deploying local Content Agentsnvidia-omniverse/content-agents checkout on branch mainNVIDIA-Omniverse/usd-convert-cad checkout for NVIDIA-backed CAD conversion
Install the individual skill from NVIDIA/skills with the skills CLI:
npx skills add nvidia/skills --skill omniverse-cad-to-simready --yes
You can target a specific agent, for example:
npx skills add nvidia/skills --skill omniverse-cad-to-simready --agent codex --yes
The source does not document the exact local directory created for the installed skill.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/omniverse-cad-to-simready" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Convert and validate only; do not deploy Content Agents.
After installation, use a request such as:
Convert /path/to/source-asset to SimReady USD in /path/to/output, including material assignment, physics assignment, validation, and optional packaging.
For conversion only, explicitly say: “Convert and validate only; do not deploy Content Agents.” Typical execution requires Python 3.12, uv, Docker, the NVIDIA Container Toolkit, an NVIDIA GPU, upstream checkouts, and the required authentication. The source directs users to references/workflow.md and references/commands.md for stage commands but does not provide the complete command list here.
What are this skill's strengths and limitations?
- Covers conversion, property assignment, conformance, validation, rendering, and packaging as an ordered workflow.
- Uses a preflight manifest to coordinate upstream dependencies, converters, and service endpoints.
- Separates basic USD viability validation from SimReady validation performed after Content Agents authoring.
- Preserves stage reports, rerun reasons, and final artifact paths for traceability and delivery.
- Local Content Agents deployment requires an NVIDIA API key, Docker, the NVIDIA Container Toolkit, and an NVIDIA GPU.
- NVIDIA-backed CAD conversion requires a separately installed and validated usd-convert-cad checkout; uv sync does not install every converter runtime.
- The documented compatibility is Linux/macOS only; Windows support is not stated.
- The skill does not ship scripts/run.py or a single workflow runner, so users must execute installed stage references.
- Grasp-physics repair requires visual evidence or explicit grasp points; otherwise the FET005 step is blocked.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| CAD to SimReady Workflow this page ✓ NVIDIA · Official | 49 · Use with care | ★ 3.8k | today | Apache-2.0 |
| Omniverse USD Performance Tuning ✓ NVIDIA · Official | 52 · Use with care | ★ 3.8k | today | Apache-2.0 |
| Optional-Dependency Smoke Tests (agent-service-toolkit) | 64 · Recommended | ★ 4.5k | 24d ago | MIT |
| skill-optimizer: Eval Workbench for Agent Skills | 61 · Recommended | ★ 82 | 5mo ago | MIT |
| Holoscan SDK Setup Guide ✓ NVIDIA · Official | 55 · Use with care | ★ 3.8k | today | Apache-2.0 |
How did FollowSkills review this skill?
The skill requires preflight, service-health checks, waiting for missing authentication, and redaction of tokens and credentials in reports; it also defines failure gates and some recovery boundaries. Points were deducted because it requires Shell, Docker, GPU resources, external services, and environment credentials without explicit per-action user confirmation, clear upload/data-flow disclosure, strong isolation, or a general rollback mechanism; an --overwrite path can be destructive.
The documentation gives a fairly detailed stage order, blockers, report fields, rerun rules, and troubleshooting guidance, making the happy path understandable statically. Points were deducted for the heavy dependency chain and many unavailable nested references and upstream checkouts; the benchmark reports an overall FAIL and duplicate implementation findings, with no statically verifiable full test coverage, so the score stays at or below the static ceiling of 10.
The intended workflow, end-to-end/conversion-only/validation-only/packaging scenarios, positive and negative evaluation cases, and browser-viewer non-fit case are reasonably clear. Points were deducted for Linux/macOS, NVIDIA GPU, Docker, and external build.nvidia.com or endpoint requirements; Chinese-language support is not described, and mainland-China network reachability is not addressed even though core cloud services may be network-constrained.
The skill includes versioning, name, tools, compatibility, license, author, tags, references, command patterns, limitations, troubleshooting, and evaluation files, with generally good structure. Points were deducted because recommended Purpose and Examples sections are absent and the description is overly long; maintenance ownership and update path are not concrete, while the benchmark explicitly reports static quality and duplication issues; repository license metadata is NOASSERTION, creating governance inconsistency with the skill's Apache-2.0 declaration.
The skill defines a clear CAD-to-SimReady orchestration, validation gates, reporting requirements, and optional packaging flow, so it plausibly covers the core task. Points were deducted because the supplied static material contains no verifiable representative final outputs or third-party execution evidence, depends on complex external services and many upstream skills, and still requires human review; the benchmark provides no Tier 3 effectiveness result and reports an overall FAIL. Under static calibration the score cannot exceed 7.
The fixed revision, supplied source files, evaluation cases, static validation summary, and specified report fields provide some auditability. Points were deducted because Tier 3 evidence is unavailable, there are no verifiable real execution results or complete committed tests covering key paths, and the benchmark reports duplication findings; only limited static evidence is available.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →