Claude Code Plugins

Community-maintained marketplace

Feedback

Structure for creating hands-on exercises in the Physical AI textbook.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name exercise-patterns
description Structure for creating hands-on exercises in the Physical AI textbook.

Exercise Template (Strict Format)

## Exercise X.Y: [Title]

**Difficulty**: [Beginner | Intermediate | Advanced]
**Time**: [15 min | 30 min | 1 hour | 2 hours]
**Hardware**: [Workstation | Jetson + RealSense | Unitree Robot]

### Objectives
By completing this exercise, you will:
- [Action verb] [specific skill] (e.g., "Create a ROS 2 publisher node")
- [Action verb] [specific skill] (e.g., "Visualize sensor data in RViz2")
- [Action verb] [specific skill] (e.g., "Deploy code to Jetson Orin Nano")

### Prerequisites
- Chapter X completed
- ROS 2 Humble installed
- [Specific hardware setup, e.g., "RealSense D435i connected"]

### Instructions

#### Step 1: [Action]
```bash
# Command to run
ros2 pkg create my_package --build-type ament_python

Expected Output:

Successfully created package 'my_package'

Step 2: [Action]

[Detailed instructions with code snippets]

Step 3: [Verification]

Run this command to verify:

ros2 topic list

Expected: You should see /my_topic in the list.

Validation Checklist

  • Code compiles without errors (colcon build)
  • Node runs and publishes data (ros2 topic echo /my_topic)
  • RViz2 displays data correctly

Challenge (Optional)

[Extended task for advanced students, e.g., "Modify the node to publish at 100 Hz instead of 10 Hz"]

Troubleshooting

Problem: "Package not found" Solution: Source your workspace (source install/setup.bash)

Problem: "Topic not visible" Solution: Check if node is running (ros2 node list)


## Exercise Types

### 1. Thought Experiment (No Code)
**Format**: Conceptual questions to build intuition
**Example**: "List 5 tasks that require physical embodiment that an LLM alone cannot do"

### 2. Simulation Task (Gazebo/Isaac Sim)
**Format**: Code + Launch files + Gazebo world
**Example**: "Spawn a humanoid in Gazebo and make it walk forward 2 meters"

### 3. Hardware Integration (Jetson + Sensors)
**Format**: Deploy ROS 2 node to Jetson, read sensor data
**Example**: "Stream RealSense depth images to your workstation via Wi-Fi"

### 4. Capstone Project (Multi-Week)
**Format**: Complete system with milestones
**Example**: "Build an autonomous room-cleaning robot"

## Progressive Difficulty Curve

**Beginner** (Chapters 1-5):
- Copy-paste code examples
- Run pre-built packages
- Simple parameter changes

**Intermediate** (Chapters 6-15):
- Modify existing code
- Create new nodes
- Integrate multiple sensors

**Advanced** (Chapters 16-28):
- Design complete systems
- Optimize for hardware (Jetson)
- Implement novel algorithms
- Deploy to real robots

## Validation Standards
Every exercise must have:
1. **Clear Success Criteria** - "You should see X" or "The robot should do Y"
2. **Runnable Code** - Copy-paste should work without modification
3. **Hardware Note** - Explicitly state if Jetson/Robot required
4. **Time Estimate** - Realistic completion time for average student