Skip to content

Latest commit

 

History

16 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MuJoCo Skill

A MuJoCo skill for AI Agent that helps AI agents handle MJCF scene construction, model checks, viewer startup, actuator inspection, and minimal control experiments more reliably.

The examples and scene checks in this README were validated with Codex GPT-5.5.

Quick Guide

1. Install the Skill

Run this in your terminal:

npx skills@latest add coolbeevip/mujoco-skills

After installation, restart Codex so the new skill metadata is reloaded.

2. Create a Simulation Scene

Enter this in Codex:

Use $mujoco to create a MuJoCo simulation scene with a franka_panda robot, a table, and a graspable object. Save it to ~/Documents/mujoco/franka_pick/scene.xml, then open the viewer to check whether the scene loads correctly.

Create simulation scene

3. Operate the Robot

Enter this in Codex:

Use $mujoco to open ~/Documents/mujoco/franka_pick/scene.xml, inspect the franka_panda actuators and grasping site, then operate the robot to pick up the object on the table.

Pick cube

More Scenarios

Use $mujoco to create a Universal Robots UR5e sorting scene with a UR5e, a conveyor belt, two cubes in different colors, and two bins. Save it to ~/Documents/mujoco/ur5e_sort/scene.xml, then open the viewer to check it.

UR5e sorting scene

Use $mujoco to create a Unitree Go1 quadruped obstacle-crossing scene with a Go1, a ground plane, low steps, and several obstacles. Save it to ~/Documents/mujoco/go1_obstacle/scene.xml, then open the viewer to check it.

Go1 obstacle scene

Use $mujoco to create a Hello Robot Stretch mobile manipulation scene with a Stretch, a tabletop, a cup, and a target tray. Save it to ~/Documents/mujoco/stretch_tabletop/scene.xml, then open the viewer to check it.

Stretch tabletop scene

Use $mujoco to create a humanoid robot balance-and-walk scene with a humanoid robot, a ground plane, low stepping blocks, and a marked walking target. Save it to ~/Documents/mujoco/humanoid_walk/scene.xml, then open the viewer to check it.

Unitree H1 humanoid walk scene

Use $mujoco to create a MuJoCo simulation of a modern manufacturing facility with production lines, industrial equipment, storage racks, narrow corridors, inspection checkpoints, and a charging pad. A Crazyflie quadrotor autonomously navigates through the factory, avoids obstacles, inspects equipment status indicators, detects anomalies, and returns to its home station.

Quadrotor autonomously scene

Use $mujoco to open ~/Documents/mujoco/factory_crazyflie_inspection/scene.xml, inspect the Crazyflie quadrotor body, actuators, cameras, and inspection waypoints, then operate the drone to take off from the home pad, fly through the factory inspection route, check both conveyor lines, inspect the central equipment area and storage racks, detect abnormal objects or blocked passages, and finally return to the home pad and land safely.

Watch the MuJoCo factory inspection video

License

MIT License.

About

A practical MuJoCo skill for AI agents to reliably create, inspect, and test MJCF robot simulation scenes.

Resources

Stars

14 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages