18. Exercises and Final Project

Beginner Exercises

Exercise 1 — AI concepts

Explain:

1. AI
2. ML
3. Deep Learning
4. Generative AI
5. LLM

Exercise 2 — Tokens

Explain why token usage matters for an AI coding agent.

Exercise 3 — Prompt

Rewrite:

Fix my application.

into a structured engineering prompt.

Intermediate Exercises

Exercise 4 — AGENTS.md

Create an AGENTS.md for a React project.

Include:

• Commands
• Architecture
• Coding rules
• Testing
• Security
• Git

Exercise 5 — Skill

Create:

.opencode/skills/api-testing/SKILL.md

The skill must define a complete API testing workflow.

Exercise 6 — Reviewer agent

Create:

.opencode/agents/reviewer.md

Requirements:

• Subagent
• No editing
• Code quality review
• Security review
• Test review

Exercise 7 — Testing agent

Create a subagent that can inspect and run tests but does not modify application code.

Advanced Exercises

Exercise 8 — Multi-agent workflow

Design:

Main
 |
 +--> Architect
 +--> Developer
 +--> Tester
 +--> Security
 +--> Reviewer

Document when each agent is used.

Exercise 9 — Permission design

Create a permission matrix for:

• Developer
• Reviewer
• Tester
• Security
• Documentation

Exercise 10 — Context optimization

Take a large repository and define:

• What belongs in AGENTS.md
• What belongs in skills
• What should remain in documentation
• What context should be retrieved per task

Final Project

Build an AI-assisted software engineering environment

Build a real application of your choice.

Examples:

• Learning management system
• QA dashboard
• Expense tracker
• College management system
• E-commerce API
• Data engineering platform

Requirements

Your project must contain:

AGENTS.md

.opencode/
|
+-- agents/
|   +-- architect.md
|   +-- reviewer.md
|   +-- tester.md
|   +-- security.md
|   +-- documentation.md
|
+-- skills/
    +-- testing/
    |   +-- SKILL.md
    |
    +-- security/
        +-- SKILL.md

Required workflow

1. Explore
2. Plan
3. Architecture
4. Implement
5. Test
6. Review
7. Security review
8. Fix
9. Document
10. Human verification

Final report

Document:

1. Architecture

What did you build?

2. AI workflow

Which agents were used?

3. Context

What was stored in AGENTS.md?

4. Skills

Which reusable skills did you create?

5. Permissions

What can each agent do?

6. Models

Which model characteristics were selected for each task and why?

7. Evaluation

How did you determine whether the AI-generated work was correct?

8. Lessons learned

What worked?

What failed?

What would you improve?

Graduation Checklist

You should be able to explain:

• [ ] AI
• [ ] Machine Learning
• [ ] Deep Learning
• [ ] Generative AI
• [ ] LLM
• [ ] Tokens
• [ ] Context window
• [ ] Prompt engineering
• [ ] Context engineering
• [ ] AI agents
• [ ] Agent loop
• [ ] Tool calling
• [ ] RAG
• [ ] Embeddings
• [ ] OpenCode
• [ ] AGENTS.md
• [ ] Skills
• [ ] SKILL.md
• [ ] Primary agents
• [ ] Subagents
• [ ] Providers
• [ ] Models
• [ ] Permissions
• [ ] MCP
• [ ] Multi-agent workflows
• [ ] AI-assisted testing
• [ ] AI-assisted code review
• [ ] AI-assisted security review

Final Principle

AI engineering is not:

Ask AI
   |
   v
Copy code

A stronger engineering workflow is:

Understand
   |
   v
Plan
   |
   v
Provide relevant context
   |
   v
Delegate
   |
   v
Implement
   |
   v
Test
   |
   v
Review
   |
   v
Verify
   |
   v
Human approval

The goal is to build software with AI while keeping engineering discipline, security and human verification.