agentleFS
NVIDIAKnown publisherNVIDIA/omniperf/.agents/skills/install-isaacsim/SKILL.md

install-isaacsim

Skill34 starsChanged 7 months agoApache-2.0View on GitHub

Install Isaac Sim via pip or source build. Covers Docker setup, verification, and common install issues. Use when the user asks to install, set up, or build Isaac Sim.

  • Deletes or force-pushes
  • Installs packages

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the required version isn't installed (Ubuntu example):

    line 43
  2. Distinctive:

    Always install Isaac Sim into an isolated Python environment.

    line 52
  3. Distinctive:

    Use this when you need to modify code, run tests, or link a custom Kit build.

    line 123
  4. Distinctive:

    Docker is required by default: Check and install if missing:

    line 127
  5. Distinctive:

    Only needed for benchmarks/scenes that use Nucleus-hosted assets.

    line 182
  6. Very specific:

    Ask the user for their OMNI_PASS token if needed

    line 190
  7. Very specific:

    Do not patch installed Isaac Sim files by default.

    line 219

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Install Isaac Sim

Public repo: https://github.com/isaac-sim/IsaacSim Branch convention: develop (latest), release/* (stable), version tags (e.g., 6.0.0)

System Requirements

  • GPU: NVIDIA RTX (Ada, Ampere, or newer recommended)
  • Driver: 535+ (check with nvidia-smi)
  • OS: Ubuntu 22.04+ (Linux), Windows 10/11
  • Python: see Python version matrix below (version is pinned per Isaac Sim release)
  • RAM: 32 GB+ recommended
  • Disk: ~30 GB for full install with cached assets

Python version matrix

Isaac Sim is built against a single, specific CPython version per release. Using the wrong Python version will cause pip install to fail to resolve wheels, or import-time ABI errors for the source build. Match the table exactly — newer or older minor versions are not supported.

Isaac SimPythonNotes
4.0.x – 4.2.x3.10Linux/Windows
4.5.x3.10Last 3.10 release
5.0.x3.11GLIBC 2.35+ required on Linux
5.1.x3.11
6.0.x3.12Current (Early Developer Release → GA)

Primary sources:

Check your Python:

python3 --version

If the required version isn't installed (Ubuntu example):

sudo add-apt-repository -y ppa:deadsnakes/ppa
sudo apt-get update
sudo apt-get install -y python3.12 python3.12-venv python3.12-dev   # adjust version to match the table

Virtual Environment (Strongly Recommended)

Always install Isaac Sim into an isolated Python environment. Installing into the system Python conflicts with distro packages (especially on Ubuntu) and makes upgrades/uninstalls messy. Use one of the options below before running any pip install command from this skill.

Pick the Python version that matches your target Isaac Sim release — see the Python version matrix. Examples below use 3.12 for Isaac Sim 6.0.x; substitute 3.11 for 5.x or 3.10 for 4.5.x.

Option A: venv (stdlib, simplest)

# Replace 3.12 with the version required by your Isaac Sim release
python3.12 -m venv ~/venvs/isaacsim
source ~/venvs/isaacsim/bin/activate
python -m pip install --upgrade pip setuptools wheel

Deactivate later with deactivate. Re-activate in any new shell before running Isaac Sim commands.

Option B: uv (fast, recommended for CI)

# Install uv once: https://docs.astral.sh/uv/
uv venv --python 3.12 ~/venvs/isaacsim   # match Python to Isaac Sim release
source ~/venvs/isaacsim/bin/activate
uv pip install --upgrade pip

Option C: conda / mamba

conda create -n isaacsim python=3.12 -y   # match Python to Isaac Sim release
conda activate isaacsim

Notes

  • Python version is strict. Each Isaac Sim release is built against one specific CPython minor version — see the matrix. Mismatches cause wheel resolution failures or import-time ABI errors.
  • Method 2 (source build) ships its own Python via _build/linux-x86_64/release/python.sh — you do not need a venv for running the source build, but you still want one for any host-side tooling (tests, scripts, editable installs).
  • Editable install (Method 3) needs an active venv — never pip install -e . into system Python.
  • Cached assets & shader caches live under ~/.cache/ov and ~/.local/share/ov. These are shared across venvs; deleting the venv does not clear them.
  • To completely reset: deactivate && rm -rf ~/venvs/isaacsim ~/.cache/ov ~/.local/share/ov.

Method 1: Pip Install (Quickest)

Note: As of April 2026, NVIDIA's pip install flow uses PyPI plus the NVIDIA package index, installs PyTorch first, and installs the full Isaac Sim extras plus the extension cache. A bare pip install isaacsim installs only the metapackage and can pass an import check while missing app, benchmark, or extension-cache components.

Activate your venv first (see Virtual Environment), then:

# Accept the Omniverse EULA non-interactively for scripts/CI.
export OMNI_KIT_ACCEPT_EULA=YES

# Isaac Sim 6.0 docs use PyTorch 2.10.0; choose the CUDA wheel index for your platform.
pip install torch==2.10.0 --index-url https://download.pytorch.org/whl/cu128

# Latest compatible full install for the active Python.
pip install 'isaacsim[all,extscache]' --extra-index-url https://pypi.nvidia.com

# Or pin to a specific release. Match the Python version matrix above.
# Isaac Sim wheels are published in 4-segment form (e.g. 4.5.0.0, 4.2.0.2);
# pin all four segments so patch releases like 4.2.0.2 are matched explicitly.
pip install 'isaacsim[all,extscache]==4.5.0.0' --extra-index-url https://pypi.nvidia.com   # Python 3.10
pip install 'isaacsim[all,extscache]==5.1.0.0' --extra-index-url https://pypi.nvidia.com   # Python 3.11
pip install 'isaacsim[all,extscache]==6.0.0.0' --extra-index-url https://pypi.nvidia.com   # Python 3.12

Verify:

python -c "import isaacsim; print('OK')"
python -c "import isaacsim; from isaacsim.simulation_app import SimulationApp; print('SimulationApp OK')"

Method 2: Source Build from GitHub

Use this when you need to modify code, run tests, or link a custom Kit build.

Prerequisites

Docker is required by default. Check and install if missing:

docker ps > /dev/null 2>&1 && echo "DOCKER OK" || echo "NEED INSTALL"

# Install Docker (Ubuntu/Debian)
sudo apt-get update && sudo apt-get install -y docker.io
sudo systemctl start docker && sudo systemctl enable docker
sudo usermod -aG docker $USER

# Start a new login session, or run this shell-only refresh:
newgrp docker

# If docker ps still fails, use `sudo docker ...` for the current command
# or ask an administrator to fix Docker group/session setup.
# Do NOT chmod /var/run/docker.sock: it grants root-equivalent Docker access
# to every local user.

Build

git clone https://github.com/isaac-sim/IsaacSim.git
cd IsaacSim
git checkout develop   # or a specific version tag

# With Docker (default, recommended)
./build.sh -xr

# Without Docker (fallback — needs extra deps)
sudo apt-get install -y libglu1-mesa-dev libegl1-mesa-dev libgles2-mesa-dev patchelf
./build.sh -xr --no-docker

Build output: _build/linux-x86_64/release

Verify source build

cd _build/linux-x86_64/release
./python.sh -c "import omni; print('OK')"

Method 3: Pip Editable Install from Source

Activate your venv first (see Virtual Environment), then:

git clone https://github.com/isaac-sim/IsaacSim.git
cd IsaacSim
git checkout develop
pip install -e .

Nucleus Authentication (Optional)

Only needed for benchmarks/scenes that use Nucleus-hosted assets. Many workflows use local assets and skip this.

export OMNI_USER='$omni-api-token'
export OMNI_PASS='<YOUR-API-TOKEN>'
  • OMNI_USER is always the literal string $omni-api-token (not a shell variable)
  • Ask the user for their OMNI_PASS token if needed
  • Expired JWT tokens cause silent failures (scenes load but materials are missing — all-black renders)

Check token expiry:

echo "$OMNI_PASS" | cut -d. -f2 | base64 -d 2>/dev/null | python3 -c "
import json, sys; from datetime import datetime; d = json.load(sys.stdin)
print(f'Expires: {datetime.fromtimestamp(d.get(\"exp\", 0))}')"

Common Issues

Build fails with Docker permission error

sudo usermod -aG docker $USER
newgrp docker
# If still fails in this shell: use `sudo docker ps` to verify Docker works,
# then start a fresh login session so group membership is applied.

No DISPLAY / Vulkan init fails

Kit needs a display even in headless mode. Set up Xvfb:

sudo apt-get install -y xvfb
Xvfb :99 -screen 0 1920x1080x24 &
export DISPLAY=:99

Shutdown hangs

Do not patch installed Isaac Sim files by default. For repeated Tracy shutdown hangs, use the scoped last-resort workaround in the profiling skill, and restore the original file afterward.


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.