AgentSkillsCN

claude-code-slash-commands

当用户希望为 Claude Code 创建、生成或构建一条斜杠命令（.md 文件）时，此方案可派上用场。此外，当用户询问斜杠命令的语法、frontmatter 选项、参数处理方式（$ARGUMENTS、$1、$2）、允许使用的工具模式、Bash 执行方式（!`command`）、文件引用（@path），或希望按命名空间组织命令时，此方案同样适用。关键词触发包括：“制作一条斜杠命令”、“为 Claude Code 创建命令”、“编写 /command”、“添加一条斜杠命令”、“命令 frontmatter”，或关于命令结构的各类疑问。

SKILL.md
--- frontmatter
name: claude-code-slash-commands
description: This skill should be used when the user wants to create, generate, or build a slash command (.md file) for Claude Code. Also use when asking about slash command syntax, frontmatter options, argument handling ($ARGUMENTS, $1, $2), allowed-tools patterns, bash execution (!`command`), file references (@path), or organizing commands by namespace. Triggers on phrases like "make a slash command", "create a command for Claude Code", "write a /command", "add a slash command", "command frontmatter", or questions about command structure.

Claude Code Slash Commands

Create custom slash commands for Claude Code as Markdown files with optional YAML frontmatter.

Quick Reference

LocationScopeDescription Label
.claude/commands/Project (shared via git)(project)
~/.claude/commands/Personal (all projects)(user)

Command Structure

markdown
---
allowed-tools: Tool1, Tool2(pattern:*)
argument-hint: [arg1] [arg2]
description: Brief description shown in /help
model: claude-sonnet-4-20250514
disable-model-invocation: false
---

Your prompt instructions here.
Use $ARGUMENTS for all args, or $1, $2, etc. for positional.
Reference files with @path/to/file.
Execute bash inline with !`command`.

Creating Commands

Workflow

  1. •Determine scope (project vs personal)
  2. •Choose command name (filename without .md)
  3. •Write frontmatter (optional but recommended)
  4. •Write prompt body with argument placeholders
  5. •Save to appropriate location

Frontmatter Options

FieldPurposeDefault
allowed-toolsTools the command can useInherits from conversation
argument-hintShows in autocomplete (e.g., [file] [options])None
descriptionBrief description for /helpFirst line of prompt
modelSpecific model to useInherits from conversation
disable-model-invocationPrevent SlashCommand tool from calling thisfalse

Arguments

All arguments - $ARGUMENTS captures everything:

markdown
Fix issue #$ARGUMENTS following our standards
# /fix-issue 123 high-priority → "123 high-priority"

Positional - $1, $2, etc. for specific args:

markdown
Review PR #$1 with priority $2 and assign to $3
# /review-pr 456 high alice → $1="456", $2="high", $3="alice"

Bash Execution

Use ! prefix to execute bash and include output in context:

markdown
---
allowed-tools: Bash(git:*)
description: Create commit from staged changes
---

## Context
- Status: !`git status --short`
- Diff: !`git diff --cached`
- Branch: !`git branch --show-current`

## Task
Create a commit message for these changes.

Important: Must include allowed-tools with Bash permissions when using ! commands.

File References

Use @ prefix to include file contents:

markdown
Review @src/utils/helpers.js for security issues
Compare @old.js with @new.js

Namespacing

Subdirectories organize commands (shown in description, not command name):

  • •.claude/commands/frontend/component.md → /component (project:frontend)
  • •~/.claude/commands/backend/api.md → /api (user:backend)

Examples

Simple Command

markdown
# .claude/commands/review.md
Review this code for bugs, security issues, and style violations.

With Arguments

markdown
---
argument-hint: [issue-number]
description: Fix GitHub issue
---
# .claude/commands/fix-issue.md
Fix issue #$ARGUMENTS following our coding standards and create a PR.

Git Commit Helper

markdown
---
allowed-tools: Bash(git add:*), Bash(git status:*), Bash(git commit:*)
argument-hint: [message]
description: Stage and commit changes
---
# .claude/commands/commit.md

## Current State
- Status: !`git status`
- Staged diff: !`git diff --cached`
- Branch: !`git branch --show-current`

## Instructions
Create a commit. If $ARGUMENTS provided, use as message. Otherwise generate from diff.

Code Generation

markdown
---
argument-hint: [component-name]
description: Generate React component
allowed-tools: Write
---
# .claude/commands/component.md

Create a React component named $1 with:
- TypeScript
- Tailwind styling
- Unit tests
- Storybook story

Reference our patterns: @src/components/Button.tsx

Multi-file Analysis

markdown
---
description: Compare implementations
argument-hint: [file1] [file2]
---
# .claude/commands/compare.md

Compare @$1 with @$2 and identify:
1. Architectural differences
2. Performance implications
3. Suggested improvements

Tips

  • •Keep descriptions concise (shown in /help)
  • •Use argument-hint for discoverability
  • •Prefer $1, $2 when args have distinct roles
  • •Use $ARGUMENTS for free-form input
  • •Test commands with various argument patterns
  • •Use extended thinking keywords for complex analysis

See Also

For complex multi-step workflows with scripts, templates, and reference docs, consider using Agent Skills instead. See references/skills-vs-commands.md.