Stage 0 — Foundations

繁體中文 | 简体中文 | English

⏱ Time estimate: 1-2 weeks (~5-15 hours, can skip if you have these)

💡 Don't recognize a term? Check resources/glossary.en.md for a 30-second definition. Stage 0 doesn't lean on much jargon, but the next stages do. 🗺️ Want the big picture of the agent landscape first (why some agents live in a terminal, some in Telegram, some on a Jetson board)? → resources/agent-paradigms.en.md (5 agent paradigms, ~10 min read)

When to skip this stage

If you can: - Write a Python function that calls a public API and parses JSON response - Use git to clone, commit, push, and resolve a basic merge - Use the command line on your OS (cd, ls, mkdir, run a script) - Read a YAML / JSON file without confusion

→ Skip directly to Stage 1.

If you can't, work through this stage. Don't skip — every later stage assumes these.

📌 Learning Goals

• Write Python: functions, classes, async/await basics
• Use git: clone, branch, commit, push, basic conflict resolution
• Use REST APIs: send GET/POST, parse JSON, handle auth headers
• Read & write YAML and JSON

🛠 Hands-on Exercises

• Exercise: Python — write a Python script that calls https://api.github.com/users/torvalds and prints follower count
• Exercise: git — clone any public repo, make a commit, push to your fork
• Exercise: CLI — make a small directory tree with the command line (macOS / Linux: mkdir project && cd project && mkdir src tests docs; Windows PowerShell: New-Item -ItemType Directory -Path project,project\src,project\tests,project\docs), run a Python script, redirect output to a file
• Exercise: YAML — read a .yaml config file in Python, modify a value, write it back
• Exercise: API auth — at github.com/settings/tokens generate a personal access token (minimal scope: read:user), call the auth-required https://api.github.com/user endpoint, observe 401 (no token) vs 200 (with token). Note: real production agents always use API auth — do this exercise

🎯 Curated Resources (not full projects, just learning material)

Python

• Python Crash Course — book + exercises (paid book, free exercises)
• Real Python tutorials — high-quality free articles
• Corey Schafer YouTube — video tutorials, beginner to advanced, very clear delivery
• Boot.dev — interactive Python course (partially free)
• runoob.com Python tutorial — Chinese-language Python intro reference

Git

• Pro Git book — free, full-length reference
• Atlassian Git Tutorials — workflow-focused
• Oh Shit, Git!?! — when things go wrong
• git-flight-rules — "I screwed up X, how do I undo?" — popular cheat sheet

CLI / Shell

• The Art of Command Line — beginner-to-advanced command-line skills (180k+ stars, multi-language)
• Learn Shell — interactive Bash tutorial
• explainshell.com — break down any shell command (debug life-saver)

REST APIs

• MDN — HTTP — protocol fundamentals
• Postman Learning Center — API exploration tool
• HTTPie — friendlier-than-curl command-line HTTP client

YAML / JSON

• YAML official site — spec
• JSON crash course — official quick guide
• jq — command-line JSON processor (heavy use in agent workflows)

Why this stage exists

Most "AI agent" tutorials assume you already have these. If you don't, you'll get blocked at random places (tools requires async; configs are YAML; SDK setup needs git). One week investing here saves 10+ weeks of frustration later.

---

✅ Done with Stage 0? Next, Stage 1 — LLM Fundamentals takes 5-8 hours to walk you through your first LLM API call, the meaning of token / context window / temperature, and how to estimate real task cost via per-token pricing. Keep going →