agentleFS
dtunaidtunai/agent-skills-for-compute/skills/omniverse-simready/SKILL.md

omniverse-simready

Skill8 starsChanged 8 months agoMITView on GitHub

NVIDIA Omniverse SimReady — OpenUSD standard for physically accurate 3D assets with metadata for robotics, digital twins, synthetic data generation

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Naming: Use clear hierarchies

    line 456
  2. Distinctive:

    Channel 1: Don't use additional channels

    line 462
  3. Distinctive:

    Pattern alignment: Before packing

    line 463
  4. Distinctive:

    Multi-component: Use multiple colliders

    line 466

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

NVIDIA Omniverse SimReady Skill

Standard and ecosystem for physically accurate 3D assets built on OpenUSD with semantic labeling, physics properties, and metadata for robotics simulation, digital twins, and synthetic data generation.

Official Sources:

What is SimReady?

Definition:

"Simulation-ready assets are physically accurate 3D objects that incorporate real-world properties, behaviors, and data bindings."

Built on OpenUSD:

  • Universal Scene Description platform
  • Modular, flexible composition
  • Cross-platform compatibility

More Than Visual:

  • Semantic labeling (WikiData integration)
  • Physics properties (USDPhysics schema)
  • Non-visual sensor attributes
  • Dense captions for context

Core Requirements

Modeling Standards

Scale:
  - Real-world scale
  - Z-up orientation
  - Export in meters
  - Pivot at origin

Geometry:
  - Optimal poly count
  - Clean topology
  - Named hierarchies
  - Forward-facing in viewport

UVs:
  - Non-overlapping islands
  - UV Channel 1 only
  - 0-1 UV space
  - 512+ pixels/meter texel density

Material Standards

Workflow:
  - PBR Metal-Rough mandatory
  - UsdPreviewSurface for portability
  - OmniPBR/OmniGlass/SimPBR for Omniverse

Textures:
  - 4K or 8K maximum resolution
  - One material per object

Shaders:
  - UsdPreviewSurface (cross-platform)
  - Omniverse MDL materials (advanced)

Physics Requirements

USDPhysics Schema:
  - Rigid body colliders
  - Mass in kilograms
  - Physical materials (friction, density, restitution)
  - Collision meshes (no convex hulls)

Compatibility:
  - PhysX
  - MuJoCo
  - Gazebo
  - Drake

Quick Start

Download SimReady Assets

# NVIDIA Physical AI Dataset
# Download from: https://developer.nvidia.com/isaac/sim

# Isaac Sim Asset Packs
# 1000+ SimReady assets (conveyors, boxes, pallets)
# Available in Isaac Sim

Load in Isaac Sim

import omni.isaac.core.utils.stage as stage_utils
from omni.isaac.core.utils.prims import create_prim

# Load SimReady asset
asset_path = "/Isaac/Environments/Simple_Warehouse/Props/S_TrafficCone.usd"
prim_path = "/World/Cone"

create_prim(
    prim_path=prim_path,
    usd_path=asset_path,
    position=[0, 0, 0]
)

Inspect SimReady Properties

from pxr import Usd, UsdPhysics, Semantics

# Open stage
stage = Usd.Stage.Open("asset.usd")
prim = stage.GetPrimAtPath("/World/Asset")

# Check physics
if prim.HasAPI(UsdPhysics.RigidBodyAPI):
    rb_api = UsdPhysics.RigidBodyAPI(prim)
    print(f"Mass: {rb_api.GetMassAttr().Get()}")

# Check semantic labels
if prim.HasAPI(Semantics.SemanticsAPI):
    sem_api = Semantics.SemanticsAPI(prim)
    labels = sem_api.GetSemanticLabelsRel().GetTargets()
    print(f"Labels: {labels}")

Semantic Labeling

WikiData Integration

# SimReady schema metadata
metadata = {
    "class": "cup",                    # Human-readable
    "hierarchy": "container/cup",      # Taxonomy
    "qcode": "Q81727",                 # WikiData unique ID
    "sku": "PROD-12345"               # Custom identifier
}

# WikiData: 100M+ objects, multi-language
# Access: https://wikidata.org
# QCode provides:
# - Synonyms
# - Taxonomic subtrees
# - Multi-language translations

Apply Semantic Labels

from pxr import Semantics, Sdf

stage = Usd.Stage.Open("asset.usd")
prim = stage.GetPrimAtPath("/World/Asset")

# Add SemanticsAPI
if not prim.HasAPI(Semantics.SemanticsAPI):
    Semantics.SemanticsAPI.Apply(prim)

sem_api = Semantics.SemanticsAPI(prim)

# Create semantic label prim
label_path = Sdf.Path("/World/Semantics/Labels/Cup")
label_prim = stage.DefinePrim(label_path, "SemanticsLabel")

# Set properties
label_prim.CreateAttribute("semantic:class", Sdf.ValueTypeNames.String).Set("cup")
label_prim.CreateAttribute("semantic:hierarchy", Sdf.ValueTypeNames.String).Set("container/cup")
label_prim.CreateAttribute("semantic:qcode", Sdf.ValueTypeNames.String).Set("Q81727")

# Link to asset
sem_api.GetSemanticLabelsRel().AddTarget(label_path)

stage.Save()

Physics Configuration

Add Rigid Body

from pxr import UsdPhysics, PhysxSchema

stage = Usd.Stage.Open("asset.usd")
root_prim = stage.GetPrimAtPath("/World/Asset")

# Apply RigidBodyAPI to root
if not root_prim.HasAPI(UsdPhysics.RigidBodyAPI):
    rb_api = UsdPhysics.RigidBodyAPI.Apply(root_prim)

# Set mass
mass_api = UsdPhysics.MassAPI.Apply(root_prim)
mass_api.GetMassAttr().Set(2.5)  # kg

stage.Save()

Add Collision Mesh

# Apply CollisionAPI to geometry
mesh_prim = stage.GetPrimAtPath("/World/Asset/Mesh")

if not mesh_prim.HasAPI(UsdPhysics.CollisionAPI):
    collision_api = UsdPhysics.CollisionAPI.Apply(mesh_prim)

# PhysX extension (optional)
if not mesh_prim.HasAPI(PhysxSchema.PhysxCollisionAPI):
    physx_collision = PhysxSchema.PhysxCollisionAPI.Apply(mesh_prim)
    physx_collision.GetContactOffsetAttr().Set(0.02)
    physx_collision.GetRestOffsetAttr().Set(0.0)

stage.Save()

Add Physical Material

# Create physics material
material_path = Sdf.Path("/World/PhysicsMaterials/Metal")
material_prim = stage.DefinePrim(material_path, "PhysicsMaterial")

# Apply PhysicsMaterialAPI
physics_mat = UsdPhysics.MaterialAPI.Apply(material_prim)
physics_mat.CreateStaticFrictionAttr().Set(0.5)
physics_mat.CreateDynamicFrictionAttr().Set(0.4)
physics_mat.CreateRestitutionAttr().Set(0.3)
physics_mat.CreateDensityAttr().Set(7850.0)  # kg/m^3 for steel

# Bind to collision mesh
binding_api = UsdShade.MaterialBindingAPI.Apply(mesh_prim)
binding_api.Bind(UsdShade.Material(material_prim),
                 bindingStrength=UsdShade.Tokens.weakerThanDescendants,
                 materialPurpose="physics")

stage.Save()

Asset Creation Workflow

1. Modeling

Requirements:
  - Real-world scale (meters)
  - Z-up orientation
  - Clean topology
  - Optimal poly count
  - Named hierarchies
  - Pivot at origin
  - Faces forward in viewport

Tools:
  - Maya
  - Blender
  - 3ds Max
  - Houdini

2. UV Layout

Standards:
  - Non-overlapping islands
  - UV Channel 1 only
  - 0-1 UV space
  - 512+ pixels/meter texel density
  - Align to material patterns
  - Re-UV CAD models

Tools:
  - RizomUV
  - Maya UV Toolkit
  - Blender UV Editor

3. Materials

Workflow: PBR Metal-Rough
Required Maps:
  - Base Color (albedo)
  - Metallic
  - Roughness
  - Normal (optional)
  - Opacity (if needed)

Resolution: 4K or 8K
Format: PNG, TIFF, or EXR

4. Physics Setup

Colliders:
  - Multiple components allowed
  - No convex hull requirement
  - Per-mesh configuration

Mass:
  - Set in kilograms
  - Real-world accurate

Materials:
  - Static friction: 0-1
  - Dynamic friction: 0-1
  - Restitution: 0-1
  - Density: kg/m^3

5. Semantic Labeling

Required Metadata:
  - Class: Human-readable name
  - Hierarchy: Taxonomy path
  - QCode: WikiData identifier
  - Custom: SKU, ProductID

Consistency: Use same taxonomy across dataset

Integration with Isaac Sim

Load SimReady Assets

from omni.isaac.core import World
from omni.isaac.core.utils.stage import add_reference_to_stage

world = World()

# Add SimReady asset
add_reference_to_stage(
    usd_path="/Isaac/Props/Pallets/pallet.usd",
    prim_path="/World/Pallet"
)

world.reset()

Synthetic Data Generation

import omni.replicator.core as rep

# Setup camera
camera = rep.create.camera(position=(5, 5, 5))

# Render with semantic labels
rp = rep.create.render_product(camera, (1024, 1024))

# Annotators
rep.AnnotatorRegistry.get_annotator("rgb")
rep.AnnotatorRegistry.get_annotator("semantic_segmentation")
rep.AnnotatorRegistry.get_annotator("bounding_box_2d_tight")
rep.AnnotatorRegistry.get_annotator("distance_to_camera")

# Randomize assets
with rep.trigger.on_frame():
    rep.randomizer.scatter_2d(
        surface_prims=["/World/Ground"],
        asset_prim="/World/Pallet"
    )

rep.orchestrator.run(num_frames=1000)

Available Asset Categories

Warehouse/Logistics:
  - Pallets (wood, plastic, metal)
  - Boxes (cardboard, plastic)
  - Conveyors
  - Shelving
  - Racks
  - Forklifts

Manufacturing:
  - Machinery
  - Tools
  - Parts bins
  - Work surfaces

Infrastructure:
  - Traffic cones
  - Barriers
  - Signs
  - Bollards

Generic Props:
  - Containers
  - Furniture
  - Equipment

Validation

Check SimReady Compliance

from pxr import Usd, UsdPhysics, Semantics

def validate_simready(asset_path):
    stage = Usd.Stage.Open(asset_path)
    root = stage.GetDefaultPrim()

    checks = {
        "has_physics": root.HasAPI(UsdPhysics.RigidBodyAPI),
        "has_mass": root.HasAPI(UsdPhysics.MassAPI),
        "has_semantics": root.HasAPI(Semantics.SemanticsAPI),
        "scale_in_meters": True,  # Manual check
        "clean_hierarchy": True   # Manual check
    }

    # Check for colliders in children
    has_collider = False
    for prim in Usd.PrimRange(root):
        if prim.HasAPI(UsdPhysics.CollisionAPI):
            has_collider = True
            break
    checks["has_collider"] = has_collider

    return checks

# Validate
results = validate_simready("asset.usd")
print(results)

Best Practices

Modeling

  1. Real-world scale: Export in meters
  2. Optimization: Balance detail vs performance
  3. Naming: Use clear hierarchies
  4. Pivot points: Position for proper placement

UVs

  1. Non-overlapping: Required for baking
  2. Texel density: Maintain consistency
  3. Channel 1: Don't use additional channels
  4. Pattern alignment: Before packing

Physics

  1. Multi-component: Use multiple colliders
  2. Mass accuracy: Real-world values
  3. Material tuning: Dial in friction/restitution
  4. Runtime compatibility: USDPhysics schema

Semantics

  1. Consistent taxonomy: Across all assets
  2. WikiData QCodes: For interoperability
  3. Hierarchical labels: Enable discovery
  4. Custom IDs: SKUs for tracking

Common Use Cases

Robotics:
  - Robot training (manipulation, navigation)
  - Sensor simulation (cameras, LIDAR)
  - Synthetic data generation
  - Sim-to-real transfer

Digital Twins:
  - Factory layouts
  - Warehouse optimization
  - Datacenter planning
  - Infrastructure design

Autonomous Vehicles:
  - Street scene simulation
  - Perception training
  - Scenario testing
  - Sensor validation

Healthcare:
  - Facility planning
  - Safety analysis
  - Equipment placement
  - Workflow optimization

References

  • Specification - SimReady standard, requirements, scope
  • Asset Creation - Modeling, UVs, materials, workflow
  • Physics Setup - USDPhysics, colliders, mass, materials
  • Semantic Labeling - WikiData, taxonomy, metadata
  • USD Structure - OpenUSD, composition, schemas
  • Isaac Sim Integration - Loading assets, synthetic data

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.