█████╗  ██████╗ ███████╗███╗   ██╗████████╗
██╔══██╗██╔════╝ ██╔════╝████╗  ██║╚══██╔══╝
███████║██║  ███╗█████╗  ██╔██╗ ██║   ██║
██╔══██║██║   ██║██╔══╝  ██║╚██╗██║   ██║
██║  ██║╚██████╔╝███████╗██║ ╚████║   ██║
╚═╝  ╚═╝ ╚═════╝ ╚══════╝╚═╝  ╚═══╝   ╚═╝
         █████╗  ██████╗ █████╗ ██████╗ ███████╗███╗   ███╗██╗   ██╗
        ██╔══██╗██╔════╝██╔══██╗██╔══██╗██╔════╝████╗ ████║╚██╗ ██╔╝
        ███████║██║     ███████║██║  ██║█████╗  ██╔████╔██║ ╚████╔╝
        ██╔══██║██║     ██╔══██║██║  ██║██╔══╝  ██║╚██╔╝██║  ╚██╔╝
        ██║  ██║╚██████╗██║  ██║██████╔╝███████╗██║ ╚═╝ ██║   ██║
        ╚═╝  ╚═╝ ╚═════╝╚═╝  ╚═╝╚═════╝ ╚══════╝╚═╝     ╚═╝   ╚═╝

TERMINAL TRAINING DIVISION

Learn to work with AI coding agents by playing a spy game in your terminal.

12 missions. Zero docs to read. One command to start.

npx agent-academy

That's it. No install. No signup. No config files. Just run it.

The Problem

AI coding agents are the most powerful coding tools most developers have never learned to use properly. The docs are thorough but dense. The features are deep but undiscoverable. You know you should learn them. You haven't yet.

The Fix

Agent Academy drops you into a spy-themed terminal training program. You're a new recruit. Your handler briefs you on each mission. You answer field assessments to prove you've internalized the material. Get it wrong and your cover takes damage. Get it right and you climb from RECRUIT to OPERATIVE to ELITE.

It's the difference between reading a manual and running a simulation.

What You'll Learn

Missions 1-4: Foundations

• What AI coding agents actually are (and aren't)
• Installation, auth, and first run
• The 6 core tools: Read, Edit, Write, Bash, Glob, Grep
• CLAUDE.md — the file that changes everything

Missions 5-8: Intermediate

• Conversation flow and slash commands
• Context window management and token economics
• The permission system (allow / ask / deny)
• Headless mode, hooks, and subagents

Missions 9-10: Advanced

• Multi-file operations and git integration
• Test-driven development with your agent
• Custom instructions and project scaffolding

Missions 11-12: Expert

• Prompt injection defense and trust boundaries
• CI/CD integration and team workflows
• Token optimization for production
• Git worktrees and parallel sessions

By mission 12, you'll know more about AI coding agents than most people who use them daily.

Features

Learn by doing, not by reading Every concept is tested with field assessments — multiple choice, command recall, and free-text evaluation. You don't move on until you prove you understand.

AI handler on standby Stuck? Press ? to ask your handler anything. It's powered by Claude and answers in the context of your current mission — not generic docs, not hallucinated examples.

Infinite practice mode Finished the curriculum? Infinite Mode generates fresh AI-powered questions across three difficulty tiers. The training never ends.

Progress that persists Your clearance level, star ratings, FXP (Field Experience Points), and achievements save between sessions. Come back tomorrow and pick up where you left off.

Spy-themed achievements From Ghost Protocol (complete all missions without contacting your handler) to Dark Hours Operative (play between 0100-0500). Some are hard. All are satisfying.

A terminal aesthetic you'll actually enjoy WarGames-meets-Hacknet visual design. Typewriter text. Pulsing animations. Double-bordered classified briefings. A Konami code easter egg. It looks like the terminal should look.

The 12 Missions

| # | Codename | Operation | What You'll Master | |---|----------|-----------|-------------------| | 1 | FIRST CONTACT | What Is Claude Code? | Agentic AI fundamentals, capabilities overview | | 2 | DEAD DROP | Installation & Setup | System requirements, auth, first launch | | 3 | BRASS COMPASS | Tools of the Trade | Read, Edit, Write, Bash, Glob, Grep | | 4 | PHANTOM PROTOCOL | The CLAUDE.md File | Persistent config, multi-level instructions | | 5 | SIGNAL CHAIN | Conversation Flow | Slash commands, prompting patterns | | 6 | GLASS HIGHWAY | Context Window | Token economics, /compact, cost control | | 7 | IRON CURTAIN | Permissions & Safety | allow/ask/deny, settings.json | | 8 | DEEP COVER | Advanced Workflows | Headless mode, hooks, subagents, piping | | 9 | DOUBLE AGENT | Multi-File & Git | Coordinated edits, TDD, Grep navigation | | 10 | LEGEND BUILDER | Custom Instructions | CLAUDE.md mastery, /init, /doctor | | 11 | BURNED NOTICE | Security | Prompt injection, .gitignore, permissions.deny, trust boundaries | | 12 | CONTROL TOWER | Production & Teams | CI/CD, token optimization, team conventions |

Optional: AI-Powered Features

The core curriculum works without any API key. For the full experience:

export ANTHROPIC_API_KEY=sk-ant-...
npx agent-academy

This unlocks:

• Live handler — ask questions mid-mission and get contextual answers
• Infinite Mode — AI-generated quizzes after you complete all 12 missions
• Free-text evaluation — answer in your own words instead of picking from options

Get your key from the Anthropic Console.

Estimated API costs: The handler and evaluator use Claude Haiku. Quiz generation uses Claude Sonnet. A full 12-mission playthrough typically costs under $0.15. Each Infinite Mode round (5 questions) costs roughly $0.02--$0.03. Actual costs depend on how often you ask your handler for help.

Tech Stack

| Layer | Technology | |-------|-----------| | Terminal UI | Ink — React for the terminal | | Components | React | | Language | TypeScript | | AI | Anthropic SDK — Claude Haiku for handler & evaluation, Sonnet for quiz generation | | Validation | Zod | | Persistence | conf — saves to ~/.config/agent-academy/ | | Testing | Vitest + ink-testing-library | | Build | tsup |

Development

git clone https://github.com/mferree/claude-code-academy.git
cd claude-code-academy
npm install
npm run dev          # watch mode
npm test             # run tests
npm run test:watch   # vitest watch mode

Contributing

Contributions are welcome! Whether it's new missions, bug fixes, accessibility improvements, or translations — we'd love your help.

License

MIT

Your training begins now.

npx agent-academy

Built with Claude Code | Powered by Anthropic