command-creator

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Command Creator

命令创建器

This skill guides the creation of Claude Code slash commands for agents in the Traycer enforcement framework. Claude Code commands are custom slash commands defined as Markdown files containing prompts.

本技能指导你在Traycer执行框架中为Agent创建Claude Code斜杠命令。Claude Code命令是包含提示词的自定义斜杠命令，以Markdown文件形式定义。

When to Use

使用场景

Use command-creator when:

Creating custom slash commands for an agent
Defining reusable prompts that should be invoked explicitly
Building commands that need dynamic arguments
Organizing frequently-used agent prompts

在以下场景使用command-creator：

为Agent创建自定义斜杠命令
定义需要显式调用的可复用提示词
构建需要动态参数的命令
整理常用的Agent提示词

Understanding Claude Code Commands

理解Claude Code命令

What Are Commands?

什么是命令？

Claude Code commands are custom slash commands that:

Are Markdown (.md) files containing prompts
Live in
```
.claude/commands/
```
(project) or
```
~/.claude/commands/
```
(personal)
Are invoked with
```
/command-name
```
syntax
Support dynamic arguments via placeholders
Can execute bash commands before running
Can reference files using
```
@
```
prefix

Commands are NOT:

Bash/CLI scripts
Python/JavaScript code
Complex workflows (those are Skills)
Automatically triggered (they require explicit invocation)

Claude Code命令是自定义斜杠命令，具备以下特性：

是包含提示词的Markdown（.md）文件
存储在项目目录的
```
.claude/commands/
```
或个人目录的
```
~/.claude/commands/
```
下
使用
```
/command-name
```
语法调用
通过占位符支持动态参数
执行前可运行bash命令
使用
```
@
```
前缀引用文件

命令不是：

Bash/CLI脚本
Python/JavaScript代码
复杂工作流（此类场景使用Skills）
自动触发的任务（需要显式调用）

Commands vs Skills vs System Prompt

命令 vs 技能 vs 系统提示词

Element	Purpose	Example	Invocation
Commands	Quick, reusable prompts	`/git-commit` , `/review-pr`	Explicit: `/command-name`
Skills	Complex workflows with multiple files	`security-validation` , `test-standards`	Automatic based on context
System Prompt	Agent personality and core behavior	Agent role, coordination logic	Always active

Use commands when:

You have a frequently-used prompt
Need explicit control over when it runs
Prompt fits in a single Markdown file
Want to pass dynamic arguments

Use skills when:

Workflow requires multiple steps
Need scripts, references, or assets
Want automatic discovery based on context
Multiple agents share the capability

Use system prompt when:

Defining agent's core role
Specifying agent-specific behavior
Coordination with other agents

元素	用途	示例	调用方式
命令	快捷、可复用的提示词	`/git-commit` , `/review-pr`	显式调用： `/command-name`
技能	涉及多文件的复杂工作流	`security-validation` , `test-standards`	根据上下文自动触发
系统提示词	定义Agent的个性和核心行为	Agent角色、协作逻辑	始终生效

优先使用命令的场景：

你有频繁使用的提示词
需要对执行时机进行显式控制
提示词可容纳在单个Markdown文件中
需要传递动态参数

优先使用技能的场景：

工作流需要多步骤执行
需要脚本、引用或资源文件
希望根据上下文自动触发
多个Agent共享该能力

优先使用系统提示词的场景：

定义Agent的核心角色
指定Agent的专属行为
与其他Agent的协作逻辑

Command Structure

命令结构

Basic Command File

基础命令文件

File:

.claude/commands/optimize.md

markdown

Analyze this code for performance issues and suggest optimizations:

Invocation:

/optimize

文件路径：

.claude/commands/optimize.md

markdown

Analyze this code for performance issues and suggest optimizations:

调用方式：

/optimize

Command with Frontmatter

包含前置元数据的命令

File:

.claude/commands/git-commit.md

markdown

---
tools:
argument-hint: [message]
description: Create a git commit
model: haiku
---

Create a git commit with message: $ARGUMENTS

Invocation:

/git-commit "fix: resolve login bug"

文件路径：

.claude/commands/git-commit.md

markdown

---
tools:
argument-hint: [message]
description: Create a git commit
model: haiku
---

Create a git commit with message: $ARGUMENTS

调用方式：

/git-commit "fix: resolve login bug"

Frontmatter Fields

前置元数据字段说明

Field	Purpose	Default	Example
`allowed-tools`	Tools command can use	Inherits from conversation	`Bash(git add:*)`
`argument-hint`	Expected arguments	None	`[pr-number] [priority]`
`description`	Brief description	First line	`Create a git commit`
`model`	Specific model	Inherits from conversation	`haiku`
`disable-model-invocation`	Prevent SlashCommand tool	false	`true`

字段	用途	默认值	示例
`allowed-tools`	命令可使用的工具	继承自对话配置	`Bash(git add:*)`
`argument-hint`	参数格式提示	无	`[pr-number] [priority]`
`description`	命令简要描述	文件第一行内容	`Create a git commit`
`model`	指定使用的模型	继承自对话配置	`haiku`
`disable-model-invocation`	禁止SlashCommand工具调用	false	`true`

Command Features

命令特性

1. Arguments

1. 参数处理

All Arguments with

$ARGUMENTS

使用

$ARGUMENTS

获取所有参数

Captures all arguments as a single string:

markdown

---
description: Fix a GitHub issue
---

Fix issue #$ARGUMENTS following our coding standards:
1. Understand the issue
2. Locate relevant code
3. Implement solution
4. Add tests
5. Prepare PR description

Usage:

bash

/fix-issue 123 high-priority

$ARGUMENTS

becomes:

"123 high-priority"

将所有参数捕获为单个字符串：

markdown

---
description: Fix a GitHub issue
---

Fix issue #$ARGUMENTS following our coding standards:
1. Understand the issue
2. Locate relevant code
3. Implement solution
4. Add tests
5. Prepare PR description

使用示例:

bash

/fix-issue 123 high-priority

$ARGUMENTS

的值为：

"123 high-priority"

Positional Arguments with

$1

$2

$3

使用

$1

$2

$3

获取位置参数

Access specific arguments individually:

markdown

---
argument-hint: [pr-number] [priority] [assignee]
description: Review pull request
---

Review PR #$1 with priority $2 and assign to $3.
Focus on security, performance, and code style.

Usage:

bash

/review-pr 456 high alice

```
$1
```
=
```
"456"
```
```
$2
```
=
```
"high"
```
```
$3
```
=
```
"alice"
```

单独访问特定位置的参数：

markdown

---
argument-hint: [pr-number] [priority] [assignee]
description: Review pull request
---

Review PR #$1 with priority $2 and assign to $3.
Focus on security, performance, and code style.

使用示例:

bash

/review-pr 456 high alice

```
$1
```
=
```
"456"
```
```
$2
```
=
```
"high"
```
```
$3
```
=
```
"alice"
```

2. Bash Command Execution

2. Bash命令执行

Execute bash commands before prompt runs using

prefix:

markdown

---
tools: Bash(git status:*), Bash(git diff:*), Bash(git log:*)
description: Create a git commit based on current changes
---

使用

前缀在提示词执行前运行bash命令：

markdown

---
tools: Bash(git status:*), Bash(git diff:*), Bash(git log:*)
description: Create a git commit based on current changes
---

Context

上下文信息

Current git status: !
```
git status
```
Staged and unstaged changes: !
```
git diff HEAD
```
Current branch: !
```
git branch --show-current
```
Recent commits: !
```
git log --oneline -10
```

当前git状态: !
```
git status
```
暂存与未暂存变更: !
```
git diff HEAD
```
当前分支: !
```
git branch --show-current
```
最近提交记录: !
```
git log --oneline -10
```

Your task

任务要求

Based on the above changes, create a single git commit with an appropriate message.


**Important**: Must include `allowed-tools` with `Bash` tool for bash execution.

基于上述变更记录，创建一个合适的git提交信息并完成提交。


**重要提示**：必须在`allowed-tools`中包含`Bash`工具才能执行bash命令。

3. File References

3. 文件引用

Include file contents using

prefix:

markdown

---
description: Review specific file implementation
---

Review the implementation in @src/utils/helpers.js and suggest improvements.

Multiple files:

markdown

Compare @src/old-version.js with @src/new-version.js and highlight differences.

使用

前缀引入文件内容：

markdown

---
description: Review specific file implementation
---

Review the implementation in @src/utils/helpers.js and suggest improvements.

多文件引用:

markdown

Compare @src/old-version.js with @src/new-version.js and highlight differences.

4. Namespacing

4. 命名空间管理

Organize commands in subdirectories:

Project structure:

text

.claude/commands/
├── git-commit.md          # /git-commit (project)
├── frontend/
│   └── component.md       # /component (project:frontend)
└── qa/
    └── review.md          # /review (project:qa)

Personal structure:

text

~/.claude/commands/
├── security-review.md     # /security-review (user)
└── templates/
    └── pr-template.md     # /pr-template (user:templates)

Conflict handling:

User vs project commands with same name: NOT SUPPORTED
Commands in different subdirectories: ALLOWED

通过子目录组织命令：

项目目录结构:

text

.claude/commands/
├── git-commit.md          # /git-commit (项目级)
├── frontend/
│   └── component.md       # /component (项目级:frontend)
└── qa/
    └── review.md          # /review (项目级:qa)

个人目录结构:

text

~/.claude/commands/
├── security-review.md     # /security-review (用户级)
└── templates/
    └── pr-template.md     # /pr-template (用户级:templates)

冲突处理:

同名的用户级与项目级命令：不支持
不同子目录下的同名命令：允许

Creating Commands

创建命令的步骤

Step 1: Determine Command Scope

步骤1：确定命令作用域

Project commands (

.claude/commands/

Shared with team via git
Project-specific workflows
Show "(project)" in
```
/help
```

Personal commands (

~/.claude/commands/

Personal workflows across all projects
Not shared with team
Show "(user)" in
```
/help
```

项目级命令（存储在

.claude/commands/

）：

通过git与团队共享
适用于项目专属工作流
在
```
/help
```
中显示"(project)"标识

用户级命令（存储在

~/.claude/commands/

）：

适用于所有项目的个人工作流
不与团队共享
在
```
/help
```
中显示"(user)"标识

Step 2: Design Command Interface

步骤2：设计命令接口

Name: Use lowercase, hyphens for spaces

Good:
```
git-commit
```
,
```
review-pr
```
,
```
fix-issue
```
Bad:
```
GitCommit
```
,
```
review_pr
```
,
```
fixIssue
```

Arguments: Decide on argument pattern

Simple: Use
```
$ARGUMENTS
```
for all args
Structured: Use
```
$1
```
,
```
$2
```
,
```
$3
```
for specific args

Prompt: Write clear instructions

Explain what the command should do
Provide context (use
```
!
```
for bash,
```
@
```
for files)
Specify expected output format

命名规则：使用小写字母，用连字符分隔单词

推荐：
```
git-commit
```
,
```
review-pr
```
,
```
fix-issue
```
不推荐：
```
GitCommit
```
,
```
review_pr
```
,
```
fixIssue
```

参数设计：确定参数模式

简单场景：使用
```
$ARGUMENTS
```
获取所有参数
结构化场景：使用
```
$1
```
,
```
$2
```
,
```
$3
```
获取特定位置参数

提示词编写：编写清晰的指令

说明命令的功能
提供上下文信息（使用
```
!
```
调用bash命令，使用
```
@
```
引用文件）
指定预期的输出格式

Step 3: Create Command File

步骤3：创建命令文件

For project command:

bash

mkdir -p .claude/commands
cat > .claude/commands/review-pr.md << 'EOF'
---
argument-hint: [pr-number]
description: Review pull request for code quality
---

Review pull request #$ARGUMENTS:

1. Check for security vulnerabilities
2. Verify test coverage
3. Assess code quality and style
4. Suggest improvements

Provide summary in Markdown format.
EOF

For personal command:

bash

mkdir -p ~/.claude/commands
cat > ~/.claude/commands/security-scan.md << 'EOF'
---
description: Scan code for security issues
---

Scan this code for security vulnerabilities:
- Hardcoded secrets
- Insecure dependencies
- Authentication issues
- Input validation problems
EOF

创建项目级命令:

bash

mkdir -p .claude/commands
cat > .claude/commands/review-pr.md << 'EOF'
---
argument-hint: [pr-number]
description: Review pull request for code quality
---

Review pull request #$ARGUMENTS:

1. Check for security vulnerabilities
2. Verify test coverage
3. Assess code quality and style
4. Suggest improvements

Provide summary in Markdown format.
EOF

创建用户级命令:

bash

mkdir -p ~/.claude/commands
cat > ~/.claude/commands/security-scan.md << 'EOF'
---
description: Scan code for security issues
---

Scan this code for security vulnerabilities:
- Hardcoded secrets
- Insecure dependencies
- Authentication issues
- Input validation problems
EOF

Step 4: Add Frontmatter

步骤4：添加前置元数据

Enhance command with metadata:

markdown

---

通过元数据增强命令功能：

markdown

---

Brief description shown in /help

在/help中显示的简要描述

description: Create a git commit

Hint shown during autocomplete

自动补全时显示的参数提示

argument-hint: [message]

Tools the command can use

命令可使用的工具

tools: Bash(git add:), Bash(git status:), Bash(git commit:*)

Specific model (optional)

指定使用的模型（可选）

model: haiku

Prevent SlashCommand tool from invoking (optional)

禁止SlashCommand工具调用（可选）

disable-model-invocation: false

Your command prompt here with $ARGUMENTS

undefined

disable-model-invocation: false

你的命令提示词，可使用$ARGUMENTS变量

undefined

Step 5: Test Command

步骤5：测试命令

bash

> /help

bash

> /help

Verify command appears in list

验证命令是否出现在列表中

/command-name arg1 arg2

Test execution with arguments

携带参数测试命令执行

undefined

undefined

Step 6: Reference in Agent Config

步骤6：在Agent配置中引用命令

File:

docs/agents/tracking-agent/config.yaml

yaml

name: tracking-agent

commands:
  - git-commit       # References .claude/commands/git-commit.md
  - linear-update    # References .claude/commands/linear-update.md
  - review-pr        # References .claude/commands/review-pr.md

Important: Reference by name without

.md

extension.

文件路径：

docs/agents/tracking-agent/config.yaml

yaml

name: tracking-agent

commands:
  - git-commit       # 引用.claude/commands/git-commit.md
  - linear-update    # 引用.claude/commands/linear-update.md
  - review-pr        # 引用.claude/commands/review-pr.md

重要提示：引用时只需使用命令名称，无需添加

.md

扩展名。

Command Examples

命令示例

Example 1: Simple Command

示例1：简单命令

File:

.claude/commands/optimize.md

markdown

Analyze the performance of this code and suggest three specific optimizations:

Usage:

/optimize

文件路径：

.claude/commands/optimize.md

markdown

Analyze the performance of this code and suggest three specific optimizations:

使用方式：

/optimize

Example 2: Command with Arguments

示例2：带参数的命令

File:

.claude/commands/fix-issue.md

markdown

---
argument-hint: [issue-number]
description: Fix a GitHub issue following project standards
---

Fix issue #$ARGUMENTS following these steps:

1. **Understand**: Read issue description and requirements
2. **Locate**: Find relevant code in codebase
3. **Implement**: Create solution addressing root cause
4. **Test**: Add appropriate tests
5. **Document**: Prepare concise PR description

Follow project coding standards and conventions.

Usage:

/fix-issue 123

文件路径：

.claude/commands/fix-issue.md

markdown

---
argument-hint: [issue-number]
description: Fix a GitHub issue following project standards
---

Fix issue #$ARGUMENTS following these steps:

1. **理解需求**：阅读问题描述和要求
2. **定位代码**：在代码库中找到相关代码
3. **实现修复**：针对根本问题创建解决方案
4. **添加测试**：编写相应的测试用例
5. **准备文档**：撰写简洁的PR描述

遵循项目的编码标准和规范。

使用方式：

/fix-issue 123

Example 3: Git Commit Command

示例3：Git提交命令

File:

.claude/commands/git-commit.md

markdown

---
tools: Bash(git add:*), Bash(git status:*), Bash(git commit:*), Bash(git diff:*), Bash(git log:*)
description: Create a git commit based on current changes
---

文件路径：

.claude/commands/git-commit.md

markdown

---
tools: Bash(git add:*), Bash(git status:*), Bash(git commit:*), Bash(git diff:*), Bash(git log:*)
description: Create a git commit based on current changes
---

Context

上下文信息

Current git status: !
```
git status
```
Staged and unstaged changes: !
```
git diff HEAD
```
Current branch: !
```
git branch --show-current
```
Recent 10 commits: !
```
git log --oneline -10
```

当前git状态: !
```
git status
```
暂存与未暂存变更: !
```
git diff HEAD
```
当前分支: !
```
git branch --show-current
```
最近10条提交记录: !
```
git log --oneline -10
```

Your Task

任务要求

Based on the above git context, create a single git commit:

Analyze changes: Review the git diff
Choose files: Select relevant files to stage

Write message: Create commit message following format:

Type:
```
feat|fix|docs|refactor|test|chore
```
Format:
```
<type>: <description>
```
Example:
```
feat: add user authentication
```

Create commit: Execute git commands to stage and commit

Important: Follow this repository's commit message style (see recent commits).


**Usage**: `/git-commit`

基于上述git上下文信息，完成以下操作：

分析变更：查看git diff记录
选择文件：选择需要暂存的相关文件

编写提交信息：按照以下格式创建提交信息：

类型:
```
feat|fix|docs|refactor|test|chore
```
格式:
```
<type>: <description>
```
示例:
```
feat: add user authentication
```

完成提交：执行git命令完成暂存和提交

重要提示：请遵循本仓库的提交信息风格（可参考最近的提交记录）。


**使用方式**：`/git-commit`

Example 4: Positional Arguments

示例4：位置参数命令

File:

.claude/commands/review-pr.md

markdown

---
argument-hint: [pr-number] [priority] [assignee]
description: Review pull request with priority and assignment
---

Review pull request #$1:

**Priority**: $2
**Assign to**: $3

**Review checklist**:
1. Security vulnerabilities
2. Performance issues
3. Code style violations
4. Test coverage
5. Documentation completeness

Provide detailed review comments with severity levels.

Usage:

/review-pr 456 high alice

文件路径：

.claude/commands/review-pr.md

markdown

---
argument-hint: [pr-number] [priority] [assignee]
description: Review pull request with priority and assignment
---

Review pull request #$1:

**优先级**: $2
**负责人**: $3

**评审检查清单**:
1. 安全漏洞
2. 性能问题
3. 代码风格违规
4. 测试覆盖率
5. 文档完整性

提供带有严重级别的详细评审意见。

使用方式：

/review-pr 456 high alice

Example 5: File Reference Command

示例5：文件引用命令

File:

.claude/commands/explain-file.md

markdown

---
argument-hint: <file-path>
description: Explain implementation of a specific file
---

Explain the implementation in @$ARGUMENTS:

1. **Purpose**: What does this file do?
2. **Structure**: How is the code organized?
3. **Key functions**: What are the main functions/classes?
4. **Dependencies**: What does it depend on?
5. **Usage**: How is it used elsewhere?

Provide clear, beginner-friendly explanation.

Usage:

/explain-file src/utils/helpers.js

文件路径：

.claude/commands/explain-file.md

markdown

---
argument-hint: <file-path>
description: Explain implementation of a specific file
---

Explain the implementation in @$ARGUMENTS:

1. **用途**：该文件的功能是什么？
2. **结构**：代码是如何组织的？
3. **核心函数**：主要的函数/类有哪些？
4. **依赖关系**：它依赖哪些其他模块？
5. **使用场景**：其他模块如何调用它？

提供清晰、适合初学者的解释。

使用方式：

/explain-file src/utils/helpers.js

Agent Config Integration

Agent配置集成

Declaring Commands

声明命令

File:

docs/agents/qa-agent/config.yaml

yaml

name: qa-agent
description: Quality assurance and validation agent

文件路径：

docs/agents/qa-agent/config.yaml

yaml

name: qa-agent
description: Quality assurance and validation agent

Skills for complex workflows

用于复杂工作流的技能

skills:

security-validation
test-standards
code-quality-standards

skills:

security-validation
test-standards
code-quality-standards

Commands for explicit invocations

用于显式调用的命令

commands:

review-pr # Quick PR review
security-scan # Security check
test-coverage # Coverage analysis

ref_docs:

test-audit-protocol.md
traycer-coordination-guide.md

undefined

commands:

review-pr # 快速PR评审
security-scan # 安全检查
test-coverage # 覆盖率分析

ref_docs:

test-audit-protocol.md
traycer-coordination-guide.md

undefined

Commands in Agent Workflow

Agent工作流中的命令

Agent can use commands:

Manually: User invokes
```
/review-pr 456
```
Programmatically: Via
```
SlashCommand
```
tool (if enabled)

SlashCommand tool requirements:

Command must have
```
description
```
frontmatter
Not disabled via
```
disable-model-invocation: true
```
Character budget not exceeded (15,000 chars default)

Agent使用命令的方式:

手动调用：用户执行
```
/review-pr 456
```
程序化调用：通过
```
SlashCommand
```
工具调用（需启用）

SlashCommand工具要求:

命令必须包含
```
description
```
前置元数据
未通过
```
disable-model-invocation: true
```
禁用
未超过字符限制（默认15000字符）

Framework Integration (Traycer Enforcement Framework)

框架集成（Traycer执行框架）

When creating commands for the Traycer enforcement framework:

为Traycer执行框架创建命令时，需遵循以下规则：

Command Location

命令存储位置

Commands live in:

.claude/commands/<command-name>.md

Example:

text

.claude/commands/
├── git-commit.md
├── review-pr.md
└── your-new-command.md

命令需存储在：

.claude/commands/<command-name>.md

示例目录结构：

text

.claude/commands/
├── git-commit.md
├── review-pr.md
└── your-new-command.md

Agent Assignment

Agent分配

When called standalone (user asks to create a command):

Ask user which agents should use this command
Suggest agents based on command's purpose:
- Git operations → tracking-agent
- Code validation → qa-agent
- Implementation → action-agent
- Workflow coordination → workflow-upgrade-assistant
List suggested agents for user confirmation

When called by agent-builder:

Agent context provided automatically
No need to ask about agent assignment

当独立调用命令创建功能时（用户请求创建命令）：

询问用户哪些Agent需要使用该命令
根据命令用途推荐合适的Agent：
- Git操作 → tracking-agent
- 代码验证 → qa-agent
- 功能实现 → action-agent
- 工作流协作 → workflow-upgrade-assistant
列出推荐的Agent供用户确认

当通过agent-builder调用时:

自动提供Agent上下文信息
无需询问Agent分配问题

Updating Agent Config

更新Agent配置

After creating the command, update each agent's config.yaml:

File:

docs/agents/<agent-name>/config.yaml

Add command name (without .md extension) to

commands:

section:

yaml

commands:
  - existing-command-1
  - existing-command-2
  - your-new-command    # Add here (no .md extension)

创建命令后，更新每个Agent的config.yaml文件：

文件路径：

docs/agents/<agent-name>/config.yaml

在

commands:

部分添加命令名称（无需

.md

扩展名）：

yaml

commands:
  - existing-command-1
  - existing-command-2
  - your-new-command    # 在此处添加（无需.md扩展名）

Complete Workflow

完整工作流

Create command file:
```
.claude/commands/<command-name>.md
```
Add frontmatter (description, argument-hint, allowed-tools)
Write command prompt with $ARGUMENTS or $1, $2, $3
Determine which agents use this command (ask user or use agent-builder context)
Update each agent's config.yaml to reference the command
Test command with
```
/command-name args
```

创建命令文件：
```
.claude/commands/<command-name>.md
```
添加前置元数据（description、argument-hint、allowed-tools）
编写命令提示词，使用$ARGUMENTS或$1, $2, $3变量
确定使用该命令的Agent（询问用户或使用agent-builder上下文）
更新每个Agent的config.yaml以引用该命令
使用
```
/command-name args
```
测试命令

Best Practices

最佳实践

1. Keep Commands Simple

1. 保持命令简洁

Good: Single-purpose, clear prompt

markdown

Review this code for security vulnerabilities

Bad: Multiple unrelated tasks

markdown

Review code for security, performance, style, tests, documentation, and deployment issues

2. Use Descriptive Names

2. 使用描述性命名

Good:

review-pr

git-commit

fix-issue

Bad:

gc

fix

3. Provide Context with Bash

3. 通过Bash提供上下文信息

Good: Include relevant git context

markdown

- Current branch: !`git branch --show-current`
- Recent commits: !`git log --oneline -5`

Bad: Assume context is known

markdown

Create a commit

4. Document Expected Arguments

4. 文档化预期参数

Good: Clear argument hint

markdown

---
argument-hint: [pr-number] [priority] [assignee]
---

Bad: No hint, unclear usage

5. Use Frontmatter

5. 使用前置元数据

Good: Complete metadata

markdown

---
description: Review pull request
argument-hint: [pr-number]
tools: Bash(gh pr view:*)
---

Bad: No frontmatter (description defaults to first line)

6. Namespace Organization

6. 命名空间组织

Good: Organized by function

text

.claude/commands/
├── git/
│   ├── commit.md
│   ├── branch.md
│   └── push.md
├── qa/
│   ├── review.md
│   └── test.md

Bad: Flat structure with many commands

Common Patterns

常见模式

Pattern 1: Context-Rich Commands

模式1：上下文丰富的命令

Include bash output for context:

markdown

---
tools: Bash(git status:*), Bash(git diff:*)
---

包含bash输出作为上下文：

markdown

---
tools: Bash(git status:*), Bash(git diff:*)
---

Context

上下文信息

Git status: !
```
git status --short
```
Changes: !
```
git diff --stat
```

Git状态: !
```
git status --short
```
变更记录: !
```
git diff --stat
```

Task

任务要求

[Your instructions based on context]

undefined

[基于上下文的指令]

undefined

Pattern 2: Structured Arguments

模式2：结构化参数

Use positional args for structured input:

markdown

---
argument-hint: [component] [action] [target]
---

$1: Component name
$2: Action to perform
$3: Target file/directory

Execute $2 on $1 targeting $3.

使用位置参数处理结构化输入：

markdown

---
argument-hint: [component] [action] [target]
---

$1: 组件名称
$2: 执行操作
$3: 目标文件/目录

对$1执行$2操作，目标为$3。

Pattern 3: File-Based Commands

模式3：基于文件的命令

Reference specific files:

markdown

Review @$ARGUMENTS for:
1. Code quality
2. Security issues
3. Performance concerns

Usage:

/review src/auth.js

引用特定文件：

markdown

Review @$ARGUMENTS检查以下内容：
1. 代码质量
2. 安全问题
3. 性能隐患

使用方式：

/review src/auth.js

Pattern 4: Multi-Step Workflows

模式4：多步骤工作流

Break complex tasks into steps:

markdown

Execute deployment workflow for $ARGUMENTS:

**Phase 1: Pre-deployment**
- [ ] Run tests
- [ ] Security scan
- [ ] Build artifacts

**Phase 2: Deployment**
- [ ] Deploy to staging
- [ ] Verify health checks
- [ ] Deploy to production

**Phase 3: Post-deployment**
- [ ] Monitor metrics
- [ ] Verify functionality
- [ ] Document deployment

将复杂任务拆分为多个步骤：

markdown

为$ARGUMENTS执行部署工作流：

**阶段1：部署前准备**
- [ ] 运行测试
- [ ] 安全扫描
- [ ] 构建产物

**阶段2：部署执行**
- [ ] 部署到预发布环境
- [ ] 验证健康检查
- [ ] 部署到生产环境

**阶段3：部署后验证**
- [ ] 监控指标
- [ ] 验证功能
- [ ] 记录部署信息

Integration with Agent Builder

与Agent Builder的集成

When using Agent Builder skill to create an agent:

Agent Builder analyzes requirements
Identifies frequently-used prompts → Candidates for commands
Calls Command Creator to define command specifications
Creates command files in
```
.claude/commands/
```
References commands in
```
config.yaml
```

Command Creator provides:

Command file structure
Frontmatter recommendations
Argument patterns
Integration with agent workflow

使用Agent Builder技能创建Agent时：

Agent Builder分析需求
识别频繁使用的提示词 → 作为命令候选
调用命令创建器定义命令规范
在.claude/commands/目录下创建命令文件
在config.yaml中引用命令

命令创建器提供以下支持:

命令文件结构
前置元数据建议
参数模式
与Agent工作流的集成方案

Quick Reference

快速参考

Creating commands:

Choose scope (project vs personal)
Design command interface (name, arguments)
Write prompt in Markdown
Add frontmatter (description, argument-hint, allowed-tools)
Test with
```
/command-name args
```
Reference in agent config.yaml

Command naming:

Use:
```
git-commit
```
,
```
review-pr
```
,
```
fix-issue
```
Avoid:
```
GitCommit
```
,
```
review_pr
```
,
```
fixIssue
```

File location:

Project:
```
.claude/commands/command-name.md
```
Personal:
```
~/.claude/commands/command-name.md
```

Config reference:

yaml

commands:
  - command-name  # No .md extension

创建命令步骤:

选择作用域（项目级 vs 用户级）
设计命令接口（名称、参数）
在Markdown中编写提示词
添加前置元数据（description、argument-hint、allowed-tools）
使用
```
/command-name args
```
测试
在Agent的config.yaml中引用

命令命名规则:

推荐:
```
git-commit
```
,
```
review-pr
```
,
```
fix-issue
```
避免:
```
GitCommit
```
,
```
review_pr
```
,
```
fixIssue
```

文件存储位置:

项目级:
```
.claude/commands/command-name.md
```
用户级:
```
~/.claude/commands/command-name.md
```

配置引用方式:

yaml

commands:
  - command-name  # 无需.md扩展名

Resources

参考资源

For detailed patterns and examples, see

references/command-examples.md

Complete command specifications from existing agents
Argument design patterns
Frontmatter best practices
Integration examples

如需查看详细模式和示例，请参考

references/command-examples.md

现有Agent的完整命令规范
参数设计模式
前置元数据最佳实践
集成示例

command-creator

Original

Translation

Command Creator

命令创建器

When to Use

使用场景

Understanding Claude Code Commands

理解Claude Code命令

What Are Commands?

什么是命令？

Commands vs Skills vs System Prompt

命令 vs 技能 vs 系统提示词

Command Structure

命令结构

Basic Command File

基础命令文件

Command with Frontmatter

包含前置元数据的命令

Frontmatter Fields

前置元数据字段说明

Command Features

命令特性

1. Arguments

1. 参数处理

All Arguments with $ARGUMENTS

使用$ARGUMENTS获取所有参数

Positional Arguments with $1, $2, $3

使用$1, $2, $3获取位置参数

2. Bash Command Execution

2. Bash命令执行

Context

上下文信息

Your task

任务要求

3. File References

3. 文件引用

4. Namespacing

4. 命名空间管理

Creating Commands

创建命令的步骤

Step 1: Determine Command Scope

步骤1：确定命令作用域

Step 2: Design Command Interface

步骤2：设计命令接口

Step 3: Create Command File

步骤3：创建命令文件

Step 4: Add Frontmatter

步骤4：添加前置元数据

Brief description shown in /help

在/help中显示的简要描述

Hint shown during autocomplete

自动补全时显示的参数提示

Tools the command can use

命令可使用的工具

Specific model (optional)

指定使用的模型（可选）

Prevent SlashCommand tool from invoking (optional)

禁止SlashCommand工具调用（可选）

disable-model-invocation: false

disable-model-invocation: false

Step 5: Test Command

步骤5：测试命令

Verify command appears in list

验证命令是否出现在列表中

Test execution with arguments

携带参数测试命令执行

Step 6: Reference in Agent Config

步骤6：在Agent配置中引用命令

Command Examples

命令示例

Example 1: Simple Command

示例1：简单命令

Example 2: Command with Arguments

示例2：带参数的命令

Example 3: Git Commit Command

示例3：Git提交命令

Context

上下文信息

Your Task

All Arguments with
`$ARGUMENTS`

使用
`$ARGUMENTS`
获取所有参数

Positional Arguments with
`$1`
,
`$2`
,
`$3`

使用
`$1`
,
`$2`
,
`$3`
获取位置参数