Claude Cursor Skill

isaac-sim

Build GPU-accelerated robot simulations and synthetic-data pipelines with NVIDIA Isaac Sim.

LLM Mart · 0 points · 1 views 0 listing impressions 0 install-command copies
Download robium-ai-robium-skills_isaac-sim-498ea4e.zip · 13 KB

Install

skills CLI npx skills add https://github.com/robium-ai/robium/tree/main/skills/isaac-sim
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install robium-ai-robium@llmmart
Git git clone https://github.com/robium-ai/robium.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole robium-ai/robium collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Isaac Sim

Isaac Sim is a gated simulation stack. Prove the target hardware and release-compatible runtime before designing the robot application around it.

Pass the gate

  • Check the current NVIDIA requirements for GPU architecture, VRAM, driver, operating system, RAM, and storage. An NVIDIA compute GPU without the required ray-tracing support is not interchangeable with a supported RTX GPU.
  • Isaac Sim does not run on macOS. Use a qualifying remote host or choose a different simulator; do not search for a local workaround.
  • Prefer an official, pinned container when local/remote parity matters. Verify the matching NGC tag, EULA requirements, host driver, and GPU passthrough at the time of use.

Build outward from a running simulation

  • Runtime: the chosen Isaac Sim release starts cleanly on the target host.
  • Scene: the USD stage, physics, units, and robot articulation match the intended application.
  • Sensors: frames, rates, ranges, noise, and rendering modes match the real interfaces or the stated synthetic-data purpose.
  • Integration: ROS 2 bridge version, distro, graph nodes, clock, and topic contracts agree with the external system.
  • Operation: batch jobs run headless; an interactive viewport uses a supported streaming path rather than display forwarding assumptions.

Go deeper only when needed

Done

  • The pinned runtime passes a small headless scene, the required sensor or ROS output is measured, and any remote viewer works through the same deployment path the application will use.
Files (robium)
  • examples
    • docker-run-command.md 3 KB
      Loading…
  • references
    • ros2-integration.md 4.8 KB
      Loading…
    • scenes-and-sensors.md 5.5 KB
      Loading…
    • setup-and-requirements.md 5.6 KB
      Loading…
    • webrtc-livestream.md 3.7 KB
      Loading…
  • FAILURES.md 2.4 KB
    Loading…
  • SKILL.md 2.4 KB
    ---
    name: isaac-sim
    description: Build GPU-accelerated robot simulations and synthetic-data pipelines with NVIDIA Isaac Sim.
    ---
    
    # Isaac Sim
    
    Isaac Sim is a gated simulation stack. Prove the target hardware and
    release-compatible runtime before designing the robot application around it.
    
    ## Pass the gate
    
    - Check the current NVIDIA requirements for GPU architecture, VRAM, driver,
      operating system, RAM, and storage. An NVIDIA compute GPU without the required
      ray-tracing support is not interchangeable with a supported RTX GPU.
    - Isaac Sim does not run on macOS. Use a qualifying remote host or choose a
      different simulator; do not search for a local workaround.
    - Prefer an official, pinned container when local/remote parity matters. Verify
      the matching NGC tag, EULA requirements, host driver, and GPU passthrough at
      the time of use.
    
    ## Build outward from a running simulation
    
    - **Runtime:** the chosen Isaac Sim release starts cleanly on the target host.
    - **Scene:** the USD stage, physics, units, and robot articulation match the
      intended application.
    - **Sensors:** frames, rates, ranges, noise, and rendering modes match the real
      interfaces or the stated synthetic-data purpose.
    - **Integration:** ROS 2 bridge version, distro, graph nodes, clock, and topic
      contracts agree with the external system.
    - **Operation:** batch jobs run headless; an interactive viewport uses a
      supported streaming path rather than display forwarding assumptions.
    
    ## Go deeper only when needed
    
    - Before installing or provisioning, read
      [references/setup-and-requirements.md](references/setup-and-requirements.md)
      and re-check its dated values against current NVIDIA documentation.
    - For USD, robot import, sensors, or Replicator, read
      [references/scenes-and-sensors.md](references/scenes-and-sensors.md).
    - For the ROS boundary, read
      [references/ros2-integration.md](references/ros2-integration.md).
    - For a remote interactive viewport, read
      [references/webrtc-livestream.md](references/webrtc-livestream.md).
    - When launch, rendering, streaming, or bridge behavior fails, start with
      [FAILURES.md](FAILURES.md).
    - Isaac Lab owns policy training. Data owns source strategy; Isaac Sim owns the
      mechanics of generating synthetic data.
    
    ## Done
    
    - The pinned runtime passes a small headless scene, the required sensor or ROS
      output is measured, and any remote viewer works through the same deployment
      path the application will use.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.