📖 Free companion courses (English) · Learn AI Coding (182 lessons) + Build AI Agents (40 lessons) — hands-on, free · also in 日本語 / Español / 한국어 · 13 courses / 648 lessons in total, all free
🌟 Sister projects: agency-agents-zh ⭐21k · superpowers-zh ⭐8.2k · agency-orchestrator ⭐2.3k · ai-shortfilm-prompts · ai-coding-trilogy · openshorts
Hands-on best practices for 10 AI coding tools — No hype, just what works. Prompt techniques, workflow design, multi-tool orchestration, and copy-paste configs to maximize your AI-assisted development.
简体中文 | English
📖 Read Online: AI Coding Trilogy → ｜ The repo content reorganized by reader persona — Volume I (Getting Started) / II (Workflows) / III (Architecture) / Appendix (Cheatsheet).
| 10 Tools Full coverage |
66 Tips Claude Code deep-dive |
7 Methodologies Prompting / Debug / Test |
Copy-paste Configs Ready to use |
Start with the 📋 Cheatsheet — all 10 tools on one page: type, context window, config files, core commands. Know what to pick, how to set it up, how to use it.
Pick a path based on your profile:
| You are... | Recommended reading order |
|---|---|
| Total newcomer to AI coding | Cheatsheet → Copilot (free tier) or Cursor (free tier) → Prompt Engineering |
| Frontend / daily coding | Cursor → Prompt Engineering → Scenarios |
| Backend / refactoring / big projects | Claude Code → Task Decomposition → Scenarios |
| Migrating from Copilot | Copilot → Claude Code (compare) → Tool Selection |
| Already on ChatGPT Plus/Pro | Codex CLI → Claude Code (compare) → Tool Selection |
| Cost-conscious / solo dev | Cheatsheet → Aider + local models or Codex CLI --oss |
| Team / high-quality delivery | Kiro → Code Review → Testing |
Already using AI tools? Jump to: Cheatsheet · Advanced Tips · Real-World Workflows · Ecosystem
| Tool | Type | Highlights |
|---|---|---|
| Claude Code | CLI Agent | 66 tips, Agent + Skill + Hook workflows |
| Codex CLI | CLI Agent | OpenAI's open-source (Rust), Sandbox + AGENTS.md, ChatGPT-plan friendly |
| Cursor | IDE | .cursor/rules/*.mdc rules, Agent / Plan modes, parallel Agents Window |
| GitHub Copilot | IDE Plugin | Inline completion + Agent mode + custom agents / instructions |
| OpenClaw | AI Agent Framework | 390k+ Stars, multi-platform + Skills + scheduled automation |
| Devin Desktop (formerly Windsurf) | IDE | Devin Local Agent, local + cloud agent board |
| Gemini CLI | CLI | By Google, 1M context; individual users have moved to Antigravity CLI |
| Kiro | IDE + CLI | By AWS, spec-driven development |
| Aider | CLI | Git-native, supports almost any LLM |
| Trae | IDE | By ByteDance, has a free tier; China edition connects directly |
Every guide follows the same structure: Core Concepts → Quick Start → Prompt Tips → Advanced Usage → Config Templates
These apply regardless of which tool you use:
| Topic | What it solves |
|---|---|
| Prompt Engineering | Prompt techniques specific to AI coding — not generic prompt engineering |
| Task Decomposition | Breaking large tasks into AI-manageable pieces |
| Code Review | Best practices for AI-assisted code review |
| Debugging | Systematic debugging methodology with AI |
| Context Management | Managing the context window to prevent AI from "getting dumber" |
| Testing Strategy | Writing tests with AI — approaches and pitfalls |
| Security | Security risks in AI-assisted coding and how to mitigate them |
How real projects run end-to-end, and how different tools play to their strengths.
| Workflow | Description |
|---|---|
| Real-World Scenarios | Refactor, collaborative dev, test backfill — 3 end-to-end scripts, copy and adapt |
| Claude Code / Cursor / Copilot / Aider — 31 real-world traps, with Symptom / Cause / Recovery / Prevention | |
| Claude Code + Cursor | Claude Code for architecture & complex refactoring, Cursor for daily coding |
| Claude Code + Copilot | Claude Code for agentic tasks, Copilot for inline completions |
| Tool Selection Guide | Which tool for which scenario — one table to rule them all |
All projects below are maintained under the「AI不止语」(AI Beyond Words) brand, covering AI-assisted development from learning to production:
Read → Learn tools → Inject methodologies → Load experts → Orchestrate → Guardrails
(book) (guide) (superpowers) (agents) (orchestrator) (shellward)
| Project | Purpose | One-liner | Setup |
|---|---|---|---|
ai-coding-trilogy  |
📚 Trilogy Book | This repo, reorganized by reader persona: Getting Started / Workflows / Architecture (PDF + web edition) | Product page |
| This repo (ai-coding-guide) | 📖 Practical Guide | 66 Claude Code tips + best practices for 10 tools + config templates | — |
superpowers-zh  |
🧠 Methodology | 20 skills teaching AI how to work (TDD / debugging / code review) | Install & Use |
agency-agents-zh  |
🎭 Expert Roles | 277 plug-and-play AI specialists, incl. 64 China-native (Xiaohongshu / Douyin / Feishu / DingTalk / Qt HMI / mechanical design) | Install & Use |
agency-orchestrator  |
🚀 Orchestration Engine | One prompt → 276 specialists collaborate, plan in minutes (15 LLMs / 11 key-free) | Install & Use |
| shellward | 🛡️ Security Middleware | 8-layer defense + DLP dataflow + injection detection, zero-dep (MCP Server included) | Install & Use |
| 🆕 openshorts | 🎥 Short-video Pipeline | Topic in, finished video out — script / TTS / subtitles / render / publish kit in one run, first video costs $0 and no API key, local-first | — |
Most AI coding resources fall into two camps: shallow "getting started" posts, or single-tool deep dives. This guide fills the gap:
- Practical over theoretical — Every tip has been tested in real projects. No "AI will change everything" filler.
- Cross-tool coverage — Compare approaches across 10 tools. Know when to use Cursor vs Claude Code vs Copilot.
- Multi-tool workflows — The biggest wins come from combining tools. We show you how.
- Copy-paste ready — Config templates, prompt snippets, and .cursorrules you can drop into your project today.
- Prompt engineering for code — Generic prompt guides don't cover the nuances of AI coding. This one does.
This project covers "how to use the 10 tools well." Want to go deeper? See Further Learning Resources — a curated list of 20+ high-quality GitHub repos / blogs / podcasts organized by topic (Prompt Engineering / MCP Ecosystem / Claude Code Deep Dive / Agent Engineering / Chinese Resources / Official Blogs).
See CHANGELOG for important updates. Latest major update: 2026-09 all 10 tools re-verified against current official docs (Windsurf → Devin Desktop, Gemini CLI individual tier retired, retired model IDs replaced, and more).
The repository core (tutorials / templates / scripts) is permanently Apache-2.0, free for individuals.
The following use cases warrant a conversation first:
| Use case | Audience |
|---|---|
| Corporate training (10+ people / series) | Team leads / HR / engineering managers |
| Paid courses / bootcamps integrating this content | Education companies / training providers |
| AI-coding consulting (tool selection / workflow design / CI integration) | Engineering leadership |
Print or commercial e-book publishing (based on book/) |
Publishers / editors |
| Speaking engagements (offline events / annual meetings / conferences) | Conference organizers |
| White-label / internal wiki integration | Large teams |
How to reach out:
- WeChat Official Account 「AI不止语」 (search
AI_BuZhiYu) — backstage message with "合作" - GitHub Issue (prefix title with
[BUSINESS]): open inquiry
Licensing: the repository core stays Apache-2.0 forever. Content under
book/uses CC BY-NC-SA 4.0 (free for individuals; commercial use requires a license).If you're an individual learning, researching, writing a blog post, or building a demo — go right ahead, no need to ask, just credit the source.
Contributions welcome! See CONTRIBUTING.en.md.
In short: add tips, fix outdated content, share workflow experiences — Issues and PRs both work.
This guide builds on excellent work from:
- claude-code-best-practice — Claude Code best practices
- awesome-cursorrules — Cursor rules collection
- awesome-copilot — Official GitHub Copilot resources
- gemini-cli-tips — Gemini CLI tips
- Everything Claude Code — Instinct scoring, AgentShield, multi-language rules
- BMAD-METHOD — Full SDLC, agent roles, multi-platform
Make AI coding tools actually work for you.