ROS Skill

Agent Skill for ROS/ROS2 robot control via rosbridge WebSocket.

Agent (LLM) → ros_cli.py → rosbridge (WebSocket :9090) → ROS/ROS2 Robot

Overview

An AI agent skill that lets agents control ROS robots through natural language. The agent reads SKILL.md, understands available commands, and executes ros_cli.py to interact with any ROS/ROS2 system via rosbridge.

Quick Start (CLI)

Use ros_cli.py directly from the command line.

# Install
pip install websocket-client

# Launch rosbridge on the robot (ROS 2)
sudo apt install ros-${ROS_DISTRO}-rosbridge-server
ros2 launch rosbridge_server rosbridge_websocket_launch.xml

# Connect
python scripts/ros_cli.py connect

# Explore
python scripts/ros_cli.py topics list
python scripts/ros_cli.py nodes list

# Move robot forward for 3 seconds
python scripts/ros_cli.py topics publish /cmd_vel geometry_msgs/Twist \
  '{"linear":{"x":1.0,"y":0,"z":0},"angular":{"x":0,"y":0,"z":0}}' --duration 3

# Read sensor data
python scripts/ros_cli.py topics subscribe /scan sensor_msgs/LaserScan --timeout 3

Quick Start (AI Agent)

ros-skill works with any AI agent that supports Agent Skills. For easy setup, we recommend OpenClaw — install ros-skill from ClawHub and talk to your robot:

• “Connect to the ROS robot”
• “Move the robot forward 1 meter”
• “What topics are available?”

See the OpenClaw tutorial for full setup and usage.

Supported Commands

All commands output JSON. See SKILL.md for quick reference and references/COMMANDS.md for full details with output examples.

How It Works

1. The agent platform loads SKILL.md into the agent’s system prompt
2. {baseDir} in commands is replaced with the actual skill installation path
3. User asks something like “move the robot forward”