# Claude Code Best Practice — Agentic Engineering Patterns

> Claude Code Best Practice is an implementation-backed playbook for agents, commands, skills, hooks, and MCP to ship repeatable Claude Code workflows.

## Install

Copy the content below into your project:

## Quick Use

1. Install / set up:
   ```bash
   git clone https://github.com/shanraisshan/claude-code-best-practice && cd claude-code-best-practice
   ```
2. Start / smoke test:
   ```bash
   ls -la .claude/commands | head
   ```
3. Verify:
   - In Claude Code, run `/weather-orchestrator` and confirm it demonstrates a command → agent → skill flow end-to-end.

## Intro

Claude Code Best Practice is an implementation-backed playbook for agents, commands, skills, hooks, and MCP to ship repeatable Claude Code workflows.

- **Best for:** Claude Code users moving from ad-hoc prompting to repeatable, reviewable agent workflows
- **Works with:** Claude Code (.claude commands/agents/skills), hooks, MCP servers, Git worktrees
- **Setup time:** 10 minutes

## Practical Notes

- Setup time ~10 minutes (clone + locate the implemented `.claude/` examples)
- One concrete demo included: `/weather-orchestrator` workflow in `.claude/commands/`
- GitHub stars + forks (verified): see Source & Thanks

Treat this repo like an internal engineering handbook:

1. Start by reading the implemented docs (not just theory).
2. Pick one pattern (e.g., commands + subagents + skills), then copy the minimal files into a real project.
3. Turn the pattern into a team habit: keep everything in version control, review changes in PRs, and require explicit guardrails for risky operations.

If you already use multiple agents in parallel, use the repo as a checklist: consistent `.claude/` structure, hooks for formatting/tests, and a small set of shared commands the whole team relies on.

### FAQ

**Q: Is this a tool or a reference repo?**
A: Primarily a reference repo with implemented examples you can copy into projects.

**Q: What should I try first?**
A: Run the `/weather-orchestrator` workflow to understand the command → agent → skill pattern.

**Q: How do I keep practices from drifting?**
A: Version-control your `.claude/` assets and review changes in PRs like code.

## Source & Thanks

> Source: https://github.com/shanraisshan/claude-code-best-practice
> License: MIT
> GitHub stars: 52,402 · forks: 5,230

---

<!-- ZH -->

## 快速使用

1. 安装 / 设置：
   ```bash
   git clone https://github.com/shanraisshan/claude-code-best-practice && cd claude-code-best-practice
   ```
2. 启动 / 冒烟测试：
   ```bash
   ls -la .claude/commands | head
   ```
3. 验证：
   - 在 Claude Code 里运行 `/weather-orchestrator`；确认能演示一次 command → agent → skill 的完整链路。

## 简介

Claude Code Best Practice 是一份可落地的实践库：围绕 agents/commands/skills/hooks/MCP 给出最佳实践与实现示例，帮助团队把零散技巧组合成可复用的工程化工作流。

- **适合谁：** 想把 Claude Code 从“随手问”升级为“可复用工作流”的团队/个人
- **可搭配：** Claude Code 的 .claude 目录体系（commands/agents/skills）、hooks 与 MCP
- **准备时间：** 10 分钟

## 实战建议

- 准备时间约 10 分钟（克隆仓库并找到 `.claude/` 的可用示例）
- 仓库自带可量化演示：`.claude/commands/` 的 `/weather-orchestrator` 工作流
- GitHub stars / forks（已核验）：见「来源与感谢」

建议把它当作“团队内部工程手册”来用：

1. 优先阅读带实现的部分（不止是概念）。
2. 先挑一个模式落地（例如 commands + subagents + skills），把最小文件集复制进真实项目。
3. 把模式固化为团队习惯：进 Git、走 PR 审查；对高风险操作（写库、删文件、发布）写清楚护栏。

如果你在并行跑多个 agent，这个仓库也适合作为自检清单：统一 `.claude/` 结构、用 hooks 保证格式/测试、沉淀少量团队共用的高频命令。

### FAQ

**这是工具还是资料库？**
答：更偏资料库：包含可复制的实现示例，适合按需搬进项目里。

**第一步建议做什么？**
答：先跑 `/weather-orchestrator`，理解 command → agent → skill 的组合方式。

**怎么避免实践漂移？**
答：把 `.claude/` 资产纳入版本管理，像代码一样走 PR 审查。

## 来源与感谢

> Source: https://github.com/shanraisshan/claude-code-best-practice
> License: MIT
> GitHub stars: 52,402 · forks: 5,230


---
Source: https://tokrepo.com/en/workflows/claude-code-best-practice-agentic-engineering-patterns
Author: Prompt Lab