前言

Claude Code 作为 Anthropic 推出的智能AI编程助手，在开发者圈子里掀起了一股热潮，给许多朋友安利试用以后，有着各种问题：“这个工具看起来很强大，但我不知道从哪里开始”、“用了几次感觉没发挥出它的潜力”、“官方文档都是英文的，看起来好费劲”… 确实，Claude Code 的学习曲线对很多开发者来说是个挑战。虽然官方发布了大量高质量的教程和视频，但由于语言障碍、访问限制等原因，很多国内开发者难以直接获取这些宝贵的学习资源。 作为一个深度使用 Claude Code 的开发者，我决定启动这个系列，将官方的核心文档翻译成中文，并结合实际使用经验进行解读，帮助大家快速上手这个强大的 AI 编程工具。 今天这篇是系列的第一篇，我们先从近期众多大佬推荐的一篇文章——Claude Code Best Practices开始。无论你是刚接触 Claude Code 的新手，还是想要提升使用效率的老手，本文都能让你快速掌握要点。 话不多说，让我们先来看个总结👇

原文地址：https://www.anthropic.com/engineering/claude-code-best-practices

中英对照译文

We recently released Claude Code, a command line tool for agentic coding. Developed as a research project, Claude Code gives Anthropic engineers and researchers a more native way to integrate Claude into their coding workflows.
我们最近发布了 Claude Code，这是一个用于智能化编程的命令行工具。作为研究项目开发，Claude Code 为 Anthropic 的工程师和研究人员提供了一种更原生的方式，将 Claude 集成到他们的编程工作流中。

Claude Code is intentionally low-level and unopinionated, providing close to raw model access without forcing specific workflows. This design philosophy creates a flexible, customizable, scriptable, and safe power tool. While powerful, this flexibility presents a learning curve for engineers new to agentic coding tools—at least until they develop their own best practices.
Claude Code 刻意设计为底层且不带倾向性，提供接近原始模型的访问权限，而不强制特定的工作流程。这种设计理念创造了一个灵活、可定制、可脚本化且安全的强大工具。虽然功能强大，但这种灵活性对初次接触智能化编程工具的工程师来说确实存在学习曲线——至少在他们形成自己的最佳实践之前是这样。

This post outlines general patterns that have proven effective, both for Anthropic’s internal teams and for external engineers using Claude Code across various codebases, languages, and environments. Nothing in this list is set in stone nor universally applicable; consider these suggestions as starting points. We encourage you to experiment and find what works best for you!
本文概述了一些已被证明有效的通用模式，这些模式适用于 Anthropic 内部团队以及在各种代码库、编程语言和环境中使用 Claude Code 的外部工程师。这份清单中的内容并非一成不变或普遍适用；请将这些建议作为起点。我们鼓励您进行实验，找到最适合自己的方法！

Looking for more detailed information? Our comprehensive documentation at claude.ai/code covers all the features mentioned in this post and provides additional examples, implementation details, and advanced techniques.
想了解更多详细信息？我们在 claude.ai/code 的文档涵盖了本文提到的所有功能，并提供了额外的示例、实现细节和高级技巧。

1. Customize your setup 自定义您的设置

Claude Code is an agentic coding assistant that automatically pulls context into prompts. This context gathering consumes time and tokens, but you can optimize it through environment tuning.
Claude Code 是一个智能编程助手，会自动将上下文拉入提示中。这种上下文收集会消耗时间和 token，但您可以通过环境调优来优化它。

a. Create CLAUDE.md files 创建 CLAUDE.md 文件

CLAUDE.md is a special file that Claude automatically pulls into context when starting a conversation. This makes it an ideal place for documenting:
CLAUDE.md 是一个特殊文件，Claude 在开始对话时会自动将其拉入上下文。这使它成为记录以下内容的理想位置：

• Common bash commands 常用的 bash 命令
• Core files and utility functions
  核心文件和工具函数
• Code style guidelines 代码风格指南
• Testing instructions 测试说明
• Repository etiquette (e.g., branch naming, merge vs. rebase, etc.)
  仓库规范（例如，分支命名、合并与变基等）
• Developer environment setup (e.g., pyenv use, which compilers work)
  开发环境设置（例如，pyenv 使用，哪些编译器有效）
• Any unexpected behaviors or warnings particular to the project
  项目特有的任何意外行为或警告
• Other information you want Claude to remember
  您希望 Claude 记住的其他信息

There’s no required format for CLAUDE.md files. We recommend keeping them concise and human-readable. For example:
CLAUDE.md 文件没有固定格式要求。我们建议保持简洁且易读。例如：

# Bash commands
- npm run build: Build the project
- npm run typecheck: Run the typechecker

# Code style
- Use ES modules (import/export) syntax, not CommonJS (require)
- Destructure imports when possible (eg. import { foo } from 'bar')

# Workflow
- Be sure to typecheck when you're done making a series of code changes
- Prefer running single tests, and not the whole test suite, for performance

You can place CLAUDE.md files in several locations:
您可以将 CLAUDE.md 文件放置在多个位置：

• The root of your repo, or wherever you run claude from (the most common usage). Name it CLAUDE.md and check it into git so that you can share it across sessions and with your team (recommended), or name it CLAUDE.local.md and .gitignore it
  仓库根目录，或运行 claude 的位置（最常见的用法）。将其命名为 CLAUDE.md 并提交到 git 中，以便在不同会话和团队成员之间共享（推荐），或将其命名为 CLAUDE.local.md 并将其加入 .gitignore
• Any parent of the directory where you run claude. This is most useful for monorepos, where you might run claude from root/foo, and have CLAUDE.md files in both root/CLAUDE.md and root/foo/CLAUDE.md. Both of these will be pulled into context automatically
  运行 claude 的目录的任何父目录。这对于单体仓库最有用，您可能从 root/foo 运行 claude，并在 root/CLAUDE.md 和 root/foo/CLAUDE.md 中都有 CLAUDE.md 文件。这两个文件都会自动被拉入上下文
• Any child of the directory where you run claude. This is the inverse of the above, and in this case, Claude will pull in CLAUDE.md files on demand when you work with files in child directories
  运行 claude 的目录的任何子目录。这与上述操作相反，在这种情况下，当您处理子目录中的文件时，Claude 会按需拉入 CLAUDE.md 文件
• Your home folder (~/.claude/CLAUDE.md), which applies it to all your claude sessions
  您的Claude的安装目录（~/.claude/CLAUDE.md），它适用于您所有的 claude 会话

When you run the /init command, Claude will automatically generate a CLAUDE.md for you.
当您运行 /init 命令时，Claude 会自动为您生成一个 CLAUDE.md。

b. Tune your CLAUDE.md files 调优您的 CLAUDE.md 文件

Your CLAUDE.md files become part of Claude’s prompts, so they should be refined like any frequently used prompt. A common mistake is adding extensive content without iterating on its effectiveness. Take time to experiment and determine what produces the best instruction following from the model.
您的 CLAUDE.md 文件会成为 Claude 提示的一部分，因此应该像任何频繁使用的提示一样进行优化。一个常见的错误是添加大量内容而不迭代其有效性。花时间进行实验，确定什么能让模型更好地遵循指令。

You can add content to your CLAUDE.md manually or press the # key to give Claude an instruction that it will automatically incorporate into the relevant CLAUDE.md. Many engineers use # frequently to document commands, files, and style guidelines while coding, then include CLAUDE.md changes in commits so team members benefit as well.
您可以手动向 CLAUDE.md 添加内容，或按 # 键给 Claude 一个指令，它会自动将其整合到相关的 CLAUDE.md 中。许多工程师在编码时频繁使用 # 来记录命令、文件和样式指南，然后在提交中包含 CLAUDE.md 的更改，以便团队成员也能受益。

At Anthropic, we occasionally run CLAUDE.md files through the prompt improver and often tune instructions (e.g. adding emphasis with “IMPORTANT” or “YOU MUST”) to improve adherence.
在 Anthropic，我们偶尔会通过提示改进器运行 CLAUDE.md 文件，并经常调整指令（例如，使用”IMPORTANT”或”YOU MUST”添加强调）以提高遵循度。

c. Curate Claude’s list of allowed tools 管理 Claude 的允许工具列表

By default, Claude Code requests permission for any action that might modify your system: file writes, many bash commands, MCP tools, etc. We designed Claude Code with this deliberately conservative approach to prioritize safety. You can customize the allowlist to permit additional tools that you know are safe, or to allow potentially unsafe tools that are easy to undo (e.g., file editing, git commit).
默认情况下，Claude Code 会对任何可能修改您系统的操作请求权限：文件写入、许多 bash 命令、MCP 工具等。我们特意采用这种保守的设计方式来优先考虑安全性。您可以自定义允许列表，以允许您知道是安全的额外工具，或允许容易撤销的潜在不安全工具（例如，文件编辑、git commit）。

There are four ways to manage allowed tools:
有四种方法管理允许的工具：

• Select “Always allow” when prompted during a session.
  在会话期间出现提示时选择”始终允许”。
• Use the /permissions command after starting Claude Code to add or remove tools from the allowlist. For example, you can add Edit to always allow file edits, Bash(git commit:*) to allow git commits, or mcp__puppeteer__puppeteer_navigate to allow navigating with the Puppeteer MCP server.
  启动 Claude Code 后使用 /permissions 命令来添加或移除允许列表中的工具。例如，您可以添加 Edit 来始终允许文件编辑，添加 Bash(git commit:*) 来允许 git 提交，或添加 mcp__puppeteer__puppeteer_navigate 来允许使用 Puppeteer MCP 服务器进行导航。
• Manually edit your .claude/settings.json or ~/.claude.json (we recommend checking the former into source control to share with your team).
  手动编辑您的 .claude/settings.json 或 ~/.claude.json（我们建议将前者提交到源代码管理中以与团队共享）。
• Use the --allowedTools CLI flag for session-specific permissions.
  使用 --allowedTools 命令行标志设置会话特定的权限。

d. If using GitHub, install the gh CLI 如果使用 GitHub，请安装 gh CLI

Claude knows how to use the gh CLI to interact with GitHub for creating issues, opening pull requests, reading comments, and more. Without gh installed, Claude can still use the GitHub API or MCP server (if you have it installed).
Claude 知道如何使用 gh CLI 与 GitHub 交互，用于创建问题、打开拉取请求、阅读评论等。如果没有安装 gh，Claude 仍然可以使用 GitHub API 或 MCP 服务器（如果您已安装）。

2. Give Claude more tools 为 Claude 提供更多工具

Claude has access to your shell environment, where you can build up sets of convenience scripts and functions for it just like you would for yourself. It can also leverage more complex tools through MCP and REST APIs.
Claude 可以访问您的 shell 环境，您可以像为自己构建一样，为它构建一组便捷的脚本和函数。它还可以通过 MCP 和 REST API 利用更复杂的工具。

a. Use Claude with bash tools 将 Claude 与 bash 工具一起使用

Claude Code inherits your bash environment, giving it access to all your tools. While Claude knows common utilities like unix tools and gh, it won’t know about your custom bash tools without instructions:
Claude Code 继承您的 bash 环境，使其能够访问您所有的工具。虽然 Claude 了解常见的实用工具，如 unix 工具和 gh，但如果没有说明，它不会知道您的自定义 bash 工具：

1. Tell Claude the tool name with usage examples
   告诉 Claude 工具名称和使用示例
2. Tell Claude to run --help to see tool documentation
   告诉 Claude 运行 --help 查看工具文档
3. Document frequently used tools in CLAUDE.md
   在 CLAUDE.md 中记录常用工具

b. Use Claude with MCP 将 Claude 与 MCP 一起使用

Claude Code functions as both an MCP server and client. As a client, it can connect to any number of MCP servers to access their tools in three ways:
Claude Code 既可作为 MCP 服务器也可作为客户端。作为客户端，它可以通过三种方式连接到任意数量的 MCP 服务器来访问其工具：

• In project config (available when running Claude Code in that directory)
  在项目配置中（在该目录中运行 Claude Code 时可用）
• In global config (available in all projects)
  在全局配置中（在所有项目中可用）
• In a checked-in .mcp.json file (available to anyone working in your codebase). For example, you can add Puppeteer and Sentry servers to your .mcp.json, so that every engineer working on your repo can use these out of the box.
  在已签入的 .mcp.json 文件中（可供在您代码库中工作的任何人使用）。例如，您可以将 Puppeteer 和 Sentry 服务器添加到您的 .mcp.json，这样在您仓库中工作的每个工程师都可以直接使用这些工具。

When working with MCP, it can also be helpful to launch Claude with the --mcp-debug flag to help identify configuration issues.
使用 MCP 时，使用 --mcp-debug 标志启动 Claude 有助于识别配置问题。

c. Use custom slash commands 使用自定义斜杠命令

For repeated workflows—debugging loops, log analysis, etc.—store prompt templates in Markdown files within the .claude/commands folder. These become available through the slash commands menu when you type /. You can check these commands into git to make them available for the rest of your team.
对于重复的工作流程——调试循环、日志分析等——将提示模板存储在 .claude/commands 文件夹中的 Markdown 文件里。当您输入 / 时，这些会通过斜杠命令菜单变得可用。您可以将这些命令提交到 git 中，让整个团队都能使用。

Custom slash commands can include the special keyword $ARGUMENTS to pass parameters from command invocation.
自定义斜杠命令可以包含特殊关键字 $ARGUMENTS 来传递命令调用的参数。

For example, here’s a slash command that you could use to automatically pull and fix a Github issue:
例如，这是一个可用于自动拉取和修复 GitHub 问题的斜杠命令：

Please analyze and fix the GitHub issue: $ARGUMENTS.

Follow these steps:

1. Use `gh issue view` to get the issue details
2. Understand the problem described in the issue
3. Search the codebase for relevant files
4. Implement the necessary changes to fix the issue
5. Write and run tests to verify the fix
6. Ensure code passes linting and type checking
7. Create a descriptive commit message
8. Push and create a PR

Remember to use the GitHub CLI (`gh`) for all GitHub-related tasks.

Putting the above content into .claude/commands/fix-github-issue.md makes it available as the /project:fix-github-issue command in Claude Code. You could then for example use /project:fix-github-issue 1234 to have Claude fix issue #1234. Similarly, you can add your own personal commands to the ~/.claude/commands folder for commands you want available in all of your sessions.
将上述内容放入 .claude/commands/fix-github-issue.md 中，使其在 Claude Code 中作为 /project:fix-github-issue 命令可用。例如，您可以使用 /project:fix-github-issue 1234 让 Claude 修复问题 # 1234。类似地，您可以将自己的个人命令添加到 ~/.claude/commands 文件夹中，以便在所有会话中使用这些命令。

3. Try common workflows 尝试常见的工作流程

Claude Code doesn’t impose a specific workflow, giving you the flexibility to use it how you want. Within the space this flexibility affords, several successful patterns for effectively using Claude Code have emerged across our community of users:
Claude Code 不强制特定的工作流程，让您可以灵活地按需使用。在这种灵活性提供的空间内，我们的用户社区已经涌现出几种有效使用 Claude Code 的成功模式：

a. Explore, plan, code, commit 探索、规划、编码、提交

This versatile workflow suits many problems:
这种通用的工作流程适合许多问题：

1. Ask Claude to read relevant files, images, or URLs, providing either general pointers (“read the file that handles logging”) or specific filenames (“read logging.py”), but explicitly tell it not to write any code just yet.
   让 Claude 读取相关的文件、图像或 URL，可以提供一般性指示（“读取处理日志的文件”）或具体文件名（“读取 logging.py”），但要明确告诉它暂时不要编写任何代码。
   1. This is the part of the workflow where you should consider strong use of subagents, especially for complex problems. Telling Claude to use subagents to verify details or investigate particular questions it might have, especially early on in a conversation or task, tends to preserve context availability without much downside in terms of lost efficiency.
      这是工作流程中应该考虑充分使用子代理的部分，特别是对于复杂问题。告诉 Claude 使用子代理来验证细节或调查它可能有的特定问题，特别是在对话或任务的早期，往往能保持上下文可用性，而不会在效率损失方面带来太多负面影响。
2. Ask Claude to make a plan for how to approach a specific problem. We recommend using the word “think” to trigger extended thinking mode, which gives Claude additional computation time to evaluate alternatives more thoroughly. These specific phrases are mapped directly to increasing levels of thinking budget in the system: “think” < “think hard” < “think harder” < “ultrathink.” Each level allocates progressively more thinking budget for Claude to use.
   让 Claude 制定解决特定问题的计划。我们建议使用”think”这个词来触发扩展思考模式，这会给 Claude 额外的计算时间来更彻底地评估各种方案。这些特定的短语直接映射到系统中递增的思考预算级别：“think” < “think hard” < “think harder” < “ultrathink”。每个级别都会为 Claude 分配越来越多的思考预算。
   1. If the results of this step seem reasonable, you can have Claude create a document or a GitHub issue with its plan so that you can reset to this spot if the implementation (step 3) isn’t what you want.
      如果这一步的结果看起来合理，您可以让 Claude 创建一个文档或 GitHub issue 来记录其计划，这样如果实现（步骤 3）不符合您的期望，您可以回退到这个步骤。
3. Ask Claude to implement its solution in code. This is also a good place to ask it to explicitly verify the reasonableness of its solution as it implements pieces of the solution.
   让 Claude 用代码实现其解决方案。这也是一个好时机，可以要求它在实现解决方案的各个部分时明确验证其合理性。
4. Ask Claude to commit the result and create a pull request. If relevant, this is also a good time to have Claude update any READMEs or changelogs with an explanation of what it just did.
   让 Claude 提交结果并创建拉取请求。如果相关，这也是让 Claude 更新任何 README 或变更日志的好时机，解释它刚刚做了什么。

Steps 1-2 are crucial—without them, Claude tends to jump straight to coding a solution. While sometimes that’s what you want, asking Claude to research and plan first significantly improves performance for problems requiring deeper thinking upfront.
步骤 1-2 至关重要——没有它们，Claude 往往会直接跳到编写解决方案。虽然有时这正是您想要的，但让 Claude 先进行研究和规划，对于需要前期深入思考的问题，可以显著提升性能。

b. Write tests, commit; code, iterate, commit 编写测试，提交；编码，迭代，提交

This is an Anthropic-favorite workflow for changes that are easily verifiable with unit, integration, or end-to-end tests. Test-driven development (TDD) becomes even more powerful with agentic coding:
这是 Anthropic 偏爱的工作流程，适用于可以通过单元测试、集成测试或端到端测试轻松验证的变更。测试驱动开发（TDD）在智能化编程中变得更加强大：

1. Ask Claude to write tests based on expected input/output pairs. Be explicit about the fact that you’re doing test-driven development so that it avoids creating mock implementations, even for functionality that doesn’t exist yet in the codebase.
   让 Claude 根据预期的输入/输出对编写测试。明确说明您正在进行测试驱动开发，这样它就能避免创建模拟实现，即使是对于代码库中尚不存在的功能。
2. Tell Claude to run the tests and confirm they fail. Explicitly telling it not to write any implementation code at this stage is often helpful.
   告诉 Claude 运行测试并确认它们失败。在此阶段明确告诉它不要编写任何实现代码通常很有帮助。
3. Ask Claude to commit the tests when you’re satisfied with them.
   当您对测试满意时，让 Claude 提交这些测试。
4. Ask Claude to write code that passes the tests, instructing it not to modify the tests. Tell Claude to keep going until all tests pass. It will usually take a few iterations for Claude to write code, run the tests, adjust the code, and run the tests again.
   让 Claude 编写通过测试的代码，并指示它不要修改测试。告诉 Claude 继续进行，直到所有测试都通过。Claude 通常需要几次迭代来编写代码、运行测试、调整代码并再次运行测试。
   1. At this stage, it can help to ask it to verify with independent subagents that the implementation isn’t overfitting to the tests
      在这个阶段，可以要求它通过独立的子代理验证实现是否过度拟合测试
5. Ask Claude to commit the code once you’re satisfied with the changes.
   当您对更改满意后，让 Claude 提交代码。

Claude performs best when it has a clear target to iterate against—a visual mock, a test case, or another kind of output. By providing expected outputs like tests, Claude can make changes, evaluate results, and incrementally improve until it succeeds.
Claude 在有明确的迭代目标时表现最佳——视觉原型、测试用例或其他类型的输出。通过提供像测试这样的预期输出，Claude 可以进行更改、评估结果并逐步改进直到成功。

c. Write code, screenshot result, iterate 编写代码，截图结果，迭代

Similar to the testing workflow, you can provide Claude with visual targets:
与测试工作流程类似，您可以向 Claude 提供视觉目标：

1. Give Claude a way to take browser screenshots (e.g., with the Puppeteer MCP server, an iOS simulator MCP server, or manually copy / paste screenshots into Claude).
   为 Claude 提供获取浏览器截图的方式（例如，通过 Puppeteer MCP 服务器、iOS 模拟器 MCP 服务器，或手动复制/粘贴截图到 Claude）。
2. Give Claude a visual mock by copying / pasting or drag-dropping an image, or giving Claude the image file path.
   通过复制/粘贴或拖放图像，或提供 Claude 图像文件路径，给 Claude 一个视觉原型。
3. Ask Claude to implement the design in code, take screenshots of the result, and iterate until its result matches the mock.
   让 Claude 用代码实现设计，对结果进行截图，并不断迭代直到其结果与原型匹配。
4. Ask Claude to commit when you’re satisfied.
   当您满意时，让 Claude 提交。

Like humans, Claude’s outputs tend to improve significantly with iteration. While the first version might be good, after 2-3 iterations it will typically look much better. Give Claude the tools to see its outputs for best results.
像人类一样，Claude 的输出往往会随着迭代显著改善。虽然第一版可能不错，但经过 2-3 次迭代后，通常会看起来好得多。为了获得最佳效果，请给 Claude 提供查看其输出的工具。

d. Safe YOLO mode 安全 YOLO 模式

Instead of supervising Claude, you can use claude --dangerously-skip-permissions to bypass all permission checks and let Claude work uninterrupted until completion. This works well for workflows like fixing lint errors or generating boilerplate code.
您可以使用 claude --dangerously-skip-permissions 绕过所有权限检查，让 Claude 不间断地工作直到完成，而不是监督 Claude。这对于修复代码检查错误或生成样板代码等工作流程很有效。

Letting Claude run arbitrary commands is risky and can result in data loss, system corruption, or even data exfiltration (e.g., via prompt injection attacks). To minimize these risks, use --dangerously-skip-permissions in a container without internet access. You can follow this reference implementation using Docker Dev Containers.
让 Claude 执行任意命令存在风险，可能导致数据丢失、系统损坏，甚至数据泄露（例如通过提示注入攻击）。为了最小化这些风险，请在没有互联网访问的容器中使用 --dangerously-skip-permissions。您可以参考使用 Docker Dev Containers 的参考实现。

e. Codebase Q&A 代码库常见问题

When onboarding to a new codebase, use Claude Code for learning and exploration. You can ask Claude the same sorts of questions you would ask another engineer on the project when pair programming. Claude can agentically search the codebase to answer general questions like:
在熟悉新代码库时，使用 Claude Code 进行学习和探索。您可以向 Claude 提出与结对编程时向项目中其他工程师询问的相同类型的问题。Claude 可以智能地搜索代码库来回答一般性问题，例如：

• How does logging work? 日志是如何工作的？
• How do I make a new API endpoint?
  如何创建新的 API 端点？
• What does async move { ... } do on line 134 of foo.rs?
  foo.rs 第 134 行的 async move { ... } 是做什么的？
• What edge cases does CustomerOnboardingFlowImpl handle?
  CustomerOnboardingFlowImpl 处理哪些边界情况？
• Why are we calling foo() instead of bar() on line 333?
  为什么我们在第 333 行调用 foo() 而不是 bar()？
• What’s the equivalent of line 334 of baz.py in Java?
  baz.py 第 334 行在 Java 中的等价代码是什么？

At Anthropic, using Claude Code in this way has become our core onboarding workflow, significantly improving ramp-up time and reducing load on other engineers. No special prompting is required! Simply ask questions, and Claude will explore the code to find answers.
在 Anthropic，以这种方式使用 Claude Code 已成为我们的核心入职流程，显著缩短了上手时间并减轻了其他工程师的负担。无需特殊的提示！只需提问，Claude 就会探索代码以找到答案。

f. Use Claude to interact with git 使用 Claude 与 git 交互

Claude can effectively handle many git operations. Many Anthropic engineers use Claude for 90%+ of our git interactions:
Claude 可以有效处理许多 git 操作。许多 Anthropic 工程师使用 Claude 进行 90% 以上的 git 交互：

• Searching git history to answer questions like “What changes made it into v1.2.3?”, “Who owns this particular feature?”, or “Why was this API designed this way?” It helps to explicitly prompt Claude to look through git history to answer queries like these.
  搜索 git 历史来回答诸如”哪些更改被纳入 v1.2.3 版本？”、“谁负责这个特定功能？“或”这个 API 为何这样设计？“等问题。明确提示 Claude 查看 git 历史来回答这类查询会很有帮助。
• Writing commit messages.Claude will look at your changes and recent history automatically to compose a message taking all the relevant context into account
  编写提交信息。Claude 会自动查看您的更改和最近的历史记录，综合所有相关上下文来撰写信息。
• Handling complex git operations like reverting files, resolving rebase conflicts, and comparing and grafting patches
  处理复杂的 git 操作，如恢复文件、解决变基冲突以及比较和移植补丁

g. Use Claude to interact with GitHub 使用 Claude 与 GitHub 交互

Claude Code can manage many GitHub interactions:
Claude Code 可以管理许多 GitHub 交互：

• Creating pull requests: Claude understands the shorthand “pr” and will generate appropriate commit messages based on the diff and surrounding context.
  创建拉取请求：Claude 理解缩写”pr”，并根据差异和周围上下文生成适当的提交信息。
• Implementing one-shot resolutions for simple code review comments: just tell it to fix comments on your PR (optionally, give it more specific instructions) and push back to the PR branch when it’s done.
  对简单的代码评审评论实现一次性解决：只需告诉它修复您的 PR 上的评论（可选地，给出更具体的指示），完成后推回 PR 分支。
• Fixing failing builds or linter warnings
  修复失败的构建或代码检查器警告
• Categorizing and triaging open issues by asking Claude to loop over open GitHub issues
  通过让 Claude 遍历开放的 GitHub 问题来对其进行分类和分级

This eliminates the need to remember gh command line syntax while automating routine tasks.
这消除了记住 gh 命令行语法的需要，同时自动化了日常任务。

h. Use Claude to work with Jupyter notebooks 使用 Claude 处理 Jupyter notebooks

Researchers and data scientists at Anthropic use Claude Code to read and write Jupyter notebooks. Claude can interpret outputs, including images, providing a fast way to explore and interact with data. There are no required prompts or workflows, but a workflow we recommend is to have Claude Code and a .ipynb file open side-by-side in VS Code.
Anthropic 的研究人员和数据科学家使用 Claude Code 来读取和编写 Jupyter notebooks。Claude 可以解释输出结果，包括图像，为探索和交互数据提供了快速的方式。没有必需的提示或工作流程，但我们推荐的工作流程是在 VS Code 中并排打开 Claude Code 和一个 .ipynb 文件。

You can also ask Claude to clean up or make aesthetic improvements to your Jupyter notebook before you show it to colleagues. Specifically telling it to make the notebook or its data visualizations “aesthetically pleasing” tends to help remind it that it’s optimizing for a human viewing experience.
您也可以让 Claude 在向同事展示之前清理或美化您的 Jupyter notebook。具体告诉它要使笔记本或其数据可视化”美观”，往往有助于提醒它正在为人类观看体验进行优化。

4. Optimize your workflow 优化您的工作流程

The suggestions below apply across all workflows:
以下建议适用于所有工作流程：

a. Be specific in your instructions 指令要具体

Claude Code’s success rate improves significantly with more specific instructions, especially on first attempts. Giving clear directions upfront reduces the need for course corrections later.
Claude Code 的成功率在指令更具体时显著提高，尤其是在初次尝试时。提前给出清晰的指示可以减少后期调整的需要。

For example: 例如：

Poor 不佳	Good 良好
add tests for foo.py 为 foo.py 添加测试	write a new test case for foo.py, covering the edge case where the user is logged out. avoid mocks 为 foo.py 编写一个新的测试用例，覆盖用户登出的边缘情况。避免使用模拟对象
why does ExecutionFactory have such a weird api? ExecutionFactory 为什么有如此奇怪的 API？	look through ExecutionFactory’s git history and summarize how its api came to be 查看 ExecutionFactory 的 git 历史记录，总结其 API 是如何形成的
add a calendar widget 添加一个日历小部件	look at how existing widgets are implemented on the home page to understand the patterns and specifically how code and interfaces are separated out. HotDogWidget.php is a good example to start with. then, follow the pattern to implement a new calendar widget that lets the user select a month and paginate forwards/backwards to pick a year. Build from scratch without libraries other than the ones already used in the rest of the codebase. 查看主页上现有小部件的实现方式，了解其模式，特别是代码和界面的分离方式。HotDogWidget.php 是一个很好的起点。然后，遵循该模式实现一个新的日历小部件，允许用户选择月份，并通过向前/向后分页选择年份。从零开始构建，仅使用代码库中已使用的库。

Claude can infer intent, but it can’t read minds. Specificity leads to better alignment with expectations.
Claude 可以推断意图，但它无法读心。具体性有助于更好地符合期望。

b. Give Claude images 向 Claude 提供图像

Claude excels with images and diagrams through several methods:
Claude 通过多种方法在图像和图表方面表现出色：

• Paste screenshots (pro tip: hit cmd+ctrl+shift+4 in macOS to screenshot to clipboard and ctrl+v to paste. Note that this is not cmd+v like you would usually use to paste on mac and does not work remotely.)
  粘贴截图（专业提示：在 macOS 中按 cmd+ctrl+shift+4 截图到剪贴板，再按 ctrl+v 粘贴。注意这不是通常在 Mac 上粘贴时使用的 cmd+v，且不支持远程操作。）
• Drag and drop images directly into the prompt input
  直接将图像拖放到提示输入框中
• Provide file paths for images
  提供图像的文件路径

This is particularly useful when working with design mocks as reference points for UI development, and visual charts for analysis and debugging. If you are not adding visuals to context, it can still be helpful to be clear with Claude about how important it is for the result to be visually appealing.
这在处理设计原型作为 UI 开发参考点以及用于分析和调试的可视化图表时特别有用。如果您没有向上下文中添加视觉元素，明确告诉 Claude 结果的视觉吸引力有多重要仍然会有帮助。

c. Mention files you want Claude to look at or work on 提及您希望 Claude 查看或处理的文件

Use tab-completion to quickly reference files or folders anywhere in your repository, helping Claude find or update the right resources.
使用 Tab 键自动补全功能快速引用仓库中任何位置的文件或文件夹，帮助 Claude 找到或更新正确的资源。

d. Give Claude URLs 向 Claude 提供 URL

Paste specific URLs alongside your prompts for Claude to fetch and read. To avoid permission prompts for the same domains (e.g., docs.foo.com), use /permissions to add domains to your allowlist.
将特定的 URL 与您的提示一起粘贴，以便 Claude 获取并阅读。为了避免对相同域名（例如 docs.foo.com）的权限提示，使用 /permissions 将域名添加到您的允许列表中。

e. Course correct early and often 尽早频繁地纠正方向

While auto-accept mode (shift+tab to toggle) lets Claude work autonomously, you’ll typically get better results by being an active collaborator and guiding Claude’s approach. You can get the best results by thoroughly explaining the task to Claude at the beginning, but you can also course correct Claude at any time.
虽然自动接受模式（使用 shift+tab 切换）让 Claude 能够自主工作，但通过积极参与并指导 Claude 的方法，您通常会获得更好的结果。通过在开始时向 Claude 详细解释任务，您可以获得最佳结果，但您也可以随时纠正 Claude 的方向。

These four tools help with course correction:
这四个工具有助于方向纠正：

• Ask Claude to make a plan before coding. Explicitly tell it not to code until you’ve confirmed its plan looks good.
  在编码前让 Claude 制定计划。明确告诉它在您确认其计划看起来不错之前不要编码。
• Press Escape to interrupt Claude during any phase (thinking, tool calls, file edits), preserving context so you can redirect or expand instructions.
  按 Escape 键可以在任何阶段（思考、工具调用、文件编辑）中断 Claude，同时保留上下文，以便您可以重新引导或扩展指令。
• Double-tap Escape to jump back in history, edit a previous prompt, and explore a different direction. You can edit the prompt and repeat until you get the result you’re looking for.
  双击 Escape 键可以跳回历史记录，编辑之前的提示，并探索不同的方向。您可以编辑提示并重复操作，直到获得您想要的结果。
• Ask Claude to undo changes, often in conjunction with option #2 to take a different approach.
  要求 Claude 撤销更改，通常与选项 #2 结合使用以采取不同的方法。

Though Claude Code occasionally solves problems perfectly on the first attempt, using these correction tools generally produces better solutions faster.
尽管 Claude Code 偶尔能一次性完美解决问题，但使用这些纠正工具通常能更快地产生更好的解决方案。

f. Use /clear to keep context focused 使用 /clear 保持上下文聚焦

During long sessions, Claude’s context window can fill with irrelevant conversation, file contents, and commands. This can reduce performance and sometimes distract Claude. Use the /clear command frequently between tasks to reset the context window.
在长时间会话中，Claude 的上下文窗口可能会被无关的对话、文件内容和命令填满。这可能会降低性能，有时还会分散 Claude 的注意力。在任务之间频繁使用 /clear 命令来重置上下文窗口。

g. Use checklists and scratchpads for complex workflows 使用清单和草稿板处理复杂工作流程

For large tasks with multiple steps or requiring exhaustive solutions—like code migrations, fixing numerous lint errors, or running complex build scripts—improve performance by having Claude use a Markdown file (or even a GitHub issue!) as a checklist and working scratchpad:
对于包含多个步骤或需要详尽解决方案的大型任务——例如代码迁移、修复大量代码检查错误或运行复杂的构建脚本——可以通过让 Claude 使用 Markdown 文件（甚至是 GitHub issue！）作为清单和工作草稿板来提高性能：

For example, to fix a large number of lint issues, you can do the following:
例如，要修复大量的代码检查问题，您可以这样做：

1. Tell Claude to run the lint command and write all resulting errors (with filenames and line numbers) to a Markdown checklist
   告诉 Claude 运行代码检查命令，并将所有结果错误（包括文件名和行号）写入 Markdown 清单
2. Instruct Claude to address each issue one by one, fixing and verifying before checking it off and moving to the next
   指导 Claude 逐一处理每个问题，在修复并验证后打勾，再进行下一个

h. Pass data into Claude 向 Claude 传递数据

Several methods exist for providing data to Claude:
有多种方法向 Claude 提供数据：

• Copy and paste directly into your prompt (most common approach)
  直接复制粘贴到您的提示中（最常见的方法）
• Pipe into Claude Code (e.g., cat foo.txt | claude), particularly useful for logs, CSVs, and large data
  通过管道传入 Claude Code（例如，cat foo.txt | claude），特别适用于日志、CSV 文件和大型数据
• Tell Claude to pull data via bash commands, MCP tools, or custom slash commands
  告诉 Claude 通过 bash 命令、MCP 工具或自定义斜杠命令获取数据
• Ask Claude to read files or fetch URLs (works for images too)
  让 Claude 读取文件或获取 URL（图像也适用）

Most sessions involve a combination of these approaches. For example, you can pipe in a log file, then tell Claude to use a tool to pull in additional context to debug the logs.
大多数会话都涉及这些方法的组合。例如，您可以通过管道输入日志文件，然后告诉 Claude 使用工具获取额外的上下文来调试日志。

5. Use headless mode to automate your infra 使用无头模式自动化您的基础设施

Claude Code includes headless mode for non-interactive contexts like CI, pre-commit hooks, build scripts, and automation. Use the -p flag with a prompt to enable headless mode, and --output-format stream-json for streaming JSON output.
Claude Code 包含无头模式，适用于非交互式环境，如 CI、预提交钩子、构建脚本和自动化。使用 -p 标志与提示来启用无头模式，使用 --output-format stream-json 进行流式 JSON 输出。

Note that headless mode does not persist between sessions. You have to trigger it each session.
请注意，无头模式不会在会话之间持续。您必须在每次会话中触发它。

a. Use Claude for issue triage 使用 Claude 进行问题分类

Headless mode can power automations triggered by GitHub events, such as when a new issue is created in your repository. For example, the public Claude Code repository uses Claude to inspect new issues as they come in and assign appropriate labels.
无头模式可以驱动由 GitHub 事件触发的自动化操作，例如当您的仓库中创建新问题时。例如，公共 Claude Code 仓库使用 Claude 来检查新创建的问题并分配适当的标签。

b. Use Claude as a linter 将 Claude 用作代码检查工具

Claude Code can provide subjective code reviews beyond what traditional linting tools detect, identifying issues like typos, stale comments, misleading function or variable names, and more.
Claude Code 可以提供超越传统代码检查工具的主观代码审查，识别诸如拼写错误、过时注释、具有误导性的函数或变量名等问题。

6. Uplevel with multi-Claude workflows 同时使用多个 Claude 来提升效率

Beyond standalone usage, some of the most powerful applications involve running multiple Claude instances in parallel:
除了独立使用外，一些最强大的应用涉及并行运行多个 Claude 实例：

a. Have one Claude write code; use another Claude to verify 让一个 Claude 编写代码；用另一个 Claude 进行验证

A simple but effective approach is to have one Claude write code while another reviews or tests it. Similar to working with multiple engineers, sometimes having separate context is beneficial:
一种简单但有效的方法是让一个 Claude 编写代码，同时另一个 Claude 进行审查或测试。类似于与多位工程师合作，有时拥有独立的上下文是有益的：

1. Use Claude to write code
   使用 Claude 编写代码
2. Run /clear or start a second Claude in another terminal
   运行 /clear 或在另一个终端启动第二个 Claude
3. Have the second Claude review the first Claude’s work
   让第二个 Claude 审查第一个 Claude 的工作
4. Start another Claude (or /clear again) to read both the code and review feedback
   启动另一个 Claude（或再次 /clear）来阅读代码和审查反馈
5. Have this Claude edit the code based on the feedback
   让这个 Claude 根据反馈编辑代码

You can do something similar with tests: have one Claude write tests, then have another Claude write code to make the tests pass. You can even have your Claude instances communicate with each other by giving them separate working scratchpads and telling them which one to write to and which one to read from.
您可以用测试来做类似的事情：让一个 Claude 编写测试，然后让另一个 Claude 编写代码使测试通过。您甚至可以让您的 Claude 实例相互通信，方法是给它们分别提供不同的工作草稿板，并告诉它们该写入哪个草稿板以及从哪个草稿板读取。

This separation often yields better results than having a single Claude handle everything.
这种分离通常比让单个 Claude 处理所有事情能获得更好的结果。

b. Have multiple checkouts of your repo 对您的仓库创建多个副本

Rather than waiting for Claude to complete each step, something many engineers at Anthropic do is:
与其等待 Claude 完成每一步，Anthropic 的许多工程师会这样做：

1. Create 3-4 git checkouts in separate folders
   在单独的文件夹中创建 3-4 个 git 副本
2. Open each folder in separate terminal tabs
   在单独的终端标签页中打开每个文件夹
3. Start Claude in each folder with different tasks
   在每个文件夹中使用不同任务启动 Claude
4. Cycle through to check progress and approve/deny permission requests
   循环检查进度并批准/拒绝权限请求

c. Use git worktrees 使用 git 工作树

This approach shines for multiple independent tasks, offering a lighter-weight alternative to multiple checkouts. Git worktrees allow you to check out multiple branches from the same repository into separate directories. Each worktree has its own working directory with isolated files, while sharing the same Git history and reflog.
这种方法在处理多个独立任务时表现出色，为多次检出提供了更轻量级的替代方案。Git 工作树允许您从同一仓库检出多个分支到不同的目录中。每个工作树拥有自己的工作目录和隔离的文件，同时共享相同的 Git 历史记录和引用日志。

Using git worktrees enables you to run multiple Claude sessions simultaneously on different parts of your project, each focused on its own independent task. For instance, you might have one Claude refactoring your authentication system while another builds a completely unrelated data visualization component. Since the tasks don’t overlap, each Claude can work at full speed without waiting for the other’s changes or dealing with merge conflicts:
使用 git 工作树使您能够在项目的不同部分同时运行多个 Claude 会话，每个会话专注于自己的独立任务。例如，您可能有一个 Claude 在重构您的认证系统，而另一个则在构建一个完全无关的数据可视化组件。由于任务不会重叠，每个 Claude 都可以全速工作，无需等待另一个的更改或处理合并冲突：

1. Create worktrees: git worktree add ../project-feature-a feature-a 创建工作树：git worktree add ../project-feature-a feature-a
2. Launch Claude in each worktree: cd ../project-feature-a && claude
   在每个工作树中启动 Claude：cd ../project-feature-a && claude
3. Create additional worktrees as needed (repeat steps 1-2 in new terminal tabs)
   根据需要创建额外的工作树（在新终端标签中重复步骤 1-2）

Some tips: 一些提示：

• Use consistent naming conventions
  使用一致的命名约定
• Maintain one terminal tab per worktree
  为每个工作树维护一个终端标签页
• If you’re using iTerm2 on Mac, set up notifications for when Claude needs attention
  如果您在 Mac 上使用 iTerm2，设置通知以便在 Claude 需要关注时收到提醒
• Use separate IDE windows for different worktrees
  为不同的工作树使用单独的 IDE 窗口
• Clean up when finished: git worktree remove ../project-feature-a
  完成时清理：git worktree remove ../project-feature-a

d. Use headless mode with a custom harness 使用带有自定义框架的无头模式

claude -p (headless mode) integrates Claude Code programmatically into larger workflows while leveraging its built-in tools and system prompt. There are two primary patterns for using headless mode:
claude -p（无头模式）通过编程将 Claude Code 集成到更大的工作流程中，同时利用其内置工具和系统提示。使用无头模式主要有两种模式：

1. Fanning out handles large migrations or analyses (e.g., analyzing sentiment in hundreds of logs or analyzing thousands of CSVs):
1. 分散并行处理大规模迁移或分析任务（例如，分析数百个日志中的情感或分析数千个 CSV 文件）：

1. Have Claude write a script to generate a task list. For example, generate a list of 2k files that need to be migrated from framework A to framework B.
   让 Claude 编写一个脚本来生成任务列表。例如，生成一个包含需要从框架 A 迁移到框架 B 的 2000 个文件的列表。
2. Loop through tasks, calling Claude programmatically for each and giving it a task and a set of tools it can use. For example: claude -p "migrate foo.py from React to Vue. When you are done, you MUST return the string OK if you succeeded, or FAIL if the task failed." --allowedTools Edit Bash(git commit:*)
   遍历任务，为每个任务以编程方式调用 Claude，并给它一个任务和一组可用的工具。例如：claude -p "migrate foo.py from React to Vue. When you are done, you MUST return the string OK if you succeeded, or FAIL if the task failed." --allowedTools Edit Bash(git commit:*)
3. Run the script several times and refine your prompt to get the desired outcome.
   多次运行脚本，并优化您的提示以获得期望的结果。

2. Pipelining integrates Claude into existing data/processing pipelines:
2. 将 Claude 集成到现有的数据/处理管道中：

1. Call claude -p "<your prompt>" --json | your_command, where your_command is the next step of your processing pipeline
   调用 claude -p "<your prompt>" --json | your_command，其中 your_command 是您处理流程的下一步
2. That’s it! JSON output (optional) can help provide structure for easier automated processing.
   就是这样！JSON 输出（可选）可以帮助提供结构，以便更轻松地进行自动化处理。

For both of these use cases, it can be helpful to use the --verbose flag for debugging the Claude invocation. We generally recommend turning verbose mode off in production for cleaner output.
对于这两种用例，使用 --verbose 标志来调试 Claude 调用可能会有所帮助。我们通常建议在生产环境中关闭详细模式，以获得更清晰的输出。

以上就是所有内容了, 如果喜欢就请帮我点个👍和♡吧!