Claude Code Slash / Commands

Slash/Commands are Claude Code interactive session'sshortcut control entry, by entering/instructions starting with /, you can quickly call functions, manage sessions, and customize workflows.

View all available commands:

/help

/helpIt will list all built-in commands, custom commands of the current project (Skills / Commands), and commands exposed by connected MCP servers.

Command types:

Command priority (when names are the same): Enterprise-level > Personal-level > Project-level > Built-in

I. Common Built-in Slash Commands

Built-in commands are core features that come with Claude Code, ready to use by direct input. Beginners should prioritize mastering these high-frequency command categories:

Basic Operation Commands

For example, we use the/helpcommand to view help information,Tabkey to switch menus,Escexit:

Session Management Commands

Context and Memory

Project and Configuration

Model and Output

Available model comparison table (2026):

Code and Tools

Integration and Extension

/pluginSubcommand details:

Statistics and Account

II. Built-in SkillsNEW

Built-in Skills are preset AI workflows bundled with Claude Code. Unlike built-in commands, they load detailed prompts and are then executed by Claude through reasoning, producing richer analysis and operation results. They are also triggered with/trigger.

III. Command Prefix Syntax

!Prefix usage examples:

> ! ls
> !git log --oneline -10
> !npm test -- --coverage
> !cat logs/error.log | tail -50

@Prefix usage examples:

> 对比 @src/auth/v1.ts 和 @src/auth/v2.ts 的实现差异
> 审查 @src/api/users.ts 中的错误处理逻辑是否完善

IV. Custom Skills (New Recommended Format)

Skills are the next-generation format for custom commands, replacing the old version.claude/commands/. Both formats work normally, but it is recommended to use Skills directly for new projects.

Skills vs Commands Comparison

Steps to Create a Skill

Step 1: Create the directory structure

Custom Skills come in two types, with different storage locations:

• Project Skill: Effective only in the current project, shared by the team → directory:.claude/skills/<名称>/
• Personal Skill: Common to all projects, for personal use → directory:~/.claude/skills/<名称>/

# 创建项目级 Skill（团队共享，纳入 git）
mkdir -p .claude/skills/optimize

# 创建个人级 Skill（跨项目通用，本地私有）
mkdir -p ~/.claude/skills/fix-bug

Step 2: Write SKILL.md

Create a newSKILL.mdfile, write configuration in YAML frontmatter, and write the prompt in the Markdown body below.nameThe field defines the trigger command name.

Example: Create a code performance optimization Skill

---
name: optimize
description: 分析代码性能瓶颈，给出优化建议。当用户提到性能问题时自动触发。
allowed-tools: Read, Grep, Glob
argument-hint: [目标文件或目录]
model: claude-sonnet-4-6
---

分析以下代码的性能瓶颈，给出具体的优化建议，优先考虑时间复杂度和内存占用：
$ARGUMENTS

When using, enter:

/optimize src/utils/data-processor.ts

Subdirectory Namespace Organization

You can.claude/skills/under it, use subdirectories to manage by function. The directory name is the command name:

.claude/skills/
├── optimize/            → /optimize
│   ├── SKILL.md
│   └── examples/        # 可选：示例文件辅助 Claude 理解
├── debug/               → /debug
│   ├── SKILL.md
│   └── scripts/         # 可选：辅助脚本
└── commit/              → /commit
    └── SKILL.md

V. Custom Slash Commands (Legacy Format, Compatible)

If you havereusable prompts(e.g., fixed code review requirements, repeated instruction templates), you can turn them into custom commands and invoke with one click.

Command file storage location:

Parameter syntax:

Core Principles

Custom commands are essentiallyMarkdown text filesThe filename is the command name, the file content is the prompt to execute, supports passing parameters and calling Bash commands.

Steps to Create Custom Commands

Step 1: Create a command storage directory

Custom commands come in two types, stored in different locations:

• Project commands: Only effective in the current project, shared by the team → Directory:.claude/commands/
• Personal commands: Common to all projects, for personal use → Directory:~/.claude/commands/

Take creating a project command as an example, first create the directory:

# 在项目根目录执行
mkdir -p .claude/commands

Step 2: Write the Markdown command file

Create a new.mdfile, the filename is the command name (e.g.,optimize.md→ command/optimize), and write the prompt in the file content.

Example: Create a code performance optimization command

# 写入提示词到文件
echo "分析这段代码的性能瓶颈，给出具体的优化建议，优先考虑时间复杂度和内存占用：" > .claude/commands/optimize.md

Now, enter in the session/optimize, then paste the code, and Claude will perform performance analysis as you requested!

Advanced Tips: Adding Parameters to Commands

Commands can have parameters, using$ARGUMENTS(capture all parameters) or$1 $2(retrieve parameters by position).

Example: Bug fix command with parameters

Create command file.claude/commands/fix-issue.md：

修复 Issue #$ARGUMENTS，要求：
1. 符合项目编码规范
2. 附上测试用例
3. 说明修复思路

Use the command:

/fix-issue 123  # $ARGUMENTS 会被替换成 "123"

VI. Frontmatter Field Reference

Custom command files (commands/*.md) and Skills (skills/*/SKILL.md) can have YAML Frontmatter added at the top for fine-grained configuration:

---
name: commit                          # Skills 中必填；命令文件中可选（默认取文件名）
allowed-tools: Bash(git add:*), Bash(git commit:*), Read, Edit
argument-hint: [commit-message]
description: 检查代码质量后提交
model: claude-haiku-4-5-20251001
context: fork
agent: general-purpose
disable-model-invocation: false
user-invocable: true
hooks:
  PreToolUse:
    - matcher: "Bash"
      hooks:
        - type: command
          command: "./scripts/validate.sh"
          once: true
---

Field Descriptions

allowed-toolsWriting Examples

Hook Trigger Timing

VII. Plugins and MCP Commands

Besides writing your own, you can also usepluginsandMCP serversto get more extended commands.

Naming format:

/mcp__<服务器名>__<命令名> [参数]

Common MCP Command Examples

View all available MCP commands:/help(MCP commands are displayed in a separate group at the bottom of the help list)

1. Plugin Commands

After installing a Claude Code plugin, plugin-specific commands are automatically added, usually in the format:

/plugin-name:command-name  # 避免命令名冲突

For example, after installing the Git plugin, there may be a/git:commitcommand that generates standardized commit messages with one click.

2. MCP Commands

MCP (Model Context Protocol) servers can turn the capabilities of external tools (such as GitHub, Jira) into slash commands, in the format:

/mcp__<服务器名>__<功能名> [参数]

Example:

/mcp__github__list_prs  # 列出 GitHub 仓库的 PR
/mcp__jira__create_issue "登录按钮失效" high  # 在 Jira 创建高优先级问题

VIII. Practical Command Template Library

Template Overview

Smart Git commit:

<!-- .claude/commands/commit.md -->
---
allowed-tools: Bash(git add:*), Bash(git status:*), Bash(git diff:*), Bash(git commit:*)
argument-hint: [commit-message]
description: 检查代码质量后提交
model: claude-haiku-4-5-20251001
---

## 暂存区内容
!`git diff --cached`

提交前检查以下问题，发现问题则报告并询问是否继续：
1. 遗留的 `console.log` / `print` 调试语句
2. 未处理的 `TODO` / `FIXME` 注释
3. 大段被注释掉的代码
4. 测试文件中的 `.only` / `.skip` 标记

若无问题，使用以下信息提交：$ARGUMENTS

PR code review:

<!-- .claude/commands/pr-review.md -->
---
allowed-tools: Read, Grep, Glob, Bash(git diff:*), Bash(git log:*)
description: 全面的 PR 代码审查
---

## 变更文件
!`git diff --name-only HEAD~1`

## 详细差异
!`git diff HEAD~1`

## 近期提交
!`git log --oneline -5`

按 Critical / Major / Minor 三级优先级输出审查结果，涵盖：
代码逻辑与可读性、安全漏洞、性能隐患（N+1 查询）、测试覆盖率、文档完整性

Smart test runner:

<!-- .claude/commands/test.md -->
---
allowed-tools: Bash, Read, Edit
argument-hint: [test-pattern]
description: 运行测试并自动修复失败用例
---

运行匹配 "$ARGUMENTS" 的测试：
1. 自动检测测试框架（Jest / pytest / Go test）
2. 运行指定模式的测试
3. 若有失败，分析根因并修复代码
4. 重新运行验证修复

不提供参数则运行全量测试并输出覆盖率报告。

Security vulnerability scan:

<!-- .claude/commands/security-scan.md -->
---
allowed-tools: Read, Grep, Glob
description: OWASP Top 10 安全漏洞扫描
model: claude-opus-4-6
---

以安全工程师视角审查代码库：

高危：SQL 注入、XSS、命令注入、硬编码密钥
中危：不安全的反序列化、过时依赖、敏感信息写入日志、CORS 配置过宽
低危：缺失安全响应头、错误信息暴露内部细节

输出格式：漏洞位置 + 危险级别 + 修复建议代码示例

API documentation generation:

<!-- .claude/commands/api-docs.md -->
---
allowed-tools: Read, Glob, Edit
argument-hint: [输出文件路径]
description: 自动生成 REST API 文档
---

扫描 @src/routes/ 和 @src/controllers/，为所有端点生成 Markdown 文档：
- 端点路径与 HTTP 方法
- 请求参数（Query / Body / Path）及类型
- 响应结构（成功与错误）
- 认证要求
- curl 调用示例

保存到：$ARGUMENTS（默认：docs/api.md）

Code refactoring:

<!-- .claude/commands/refactor.md -->
---
allowed-tools: Read, Edit, Glob
argument-hint: [目标文件或目录]
description: 代码质量重构
---

对 $ARGUMENTS 进行重构，遵循原则：
- 可读性：变量/函数命名语义化，消除魔法数字
- DRY 原则：提取重复逻辑为复用函数
- 单一职责：拆分超过 50 行的函数
- 错误处理：完善异常捕获和错误边界
- TypeScript：消除 `any` 类型，补全类型注解

重构前先输出变更计划，确认后再执行。

Documentation link checker:

<!-- .claude/commands/check-links.md -->
---
allowed-tools: Read, Bash, WebFetch
argument-hint: [文档文件路径]
description: 检查文档中的链接是否全部有效
model: claude-haiku-4-5-20251001
---

读取 $ARGUMENTS 文件，提取所有超链接（http/https），逐一请求并检查响应状态码。

输出格式：
- 有效链接（200）
- 重定向链接（3xx）及目标地址
- 失效链接（4xx/5xx）及建议替代链接

最后输出汇总：共 N 个链接，M 个失效。

Appendix: Configuration File Description