Create Slash Command Skill

This Skill helps you create custom slash commands for Claude Code. Slash commands are user-invoked shortcuts stored as Markdown files that can include arguments, bash execution, and file references.

Instructions

When creating a slash command, follow these steps:

1. Determine the command scope:

   • Project commands: .claude/commands/ (shared via git)
   • Personal commands: ~/.claude/commands/ (local only)

2. Ask the user for command details:

   • Command name (use lowercase, hyphens, descriptive)
   • What the command should do
   • Whether it needs arguments
   • Whether it needs to execute bash commands
   • Whether it needs to reference files

3. Create the command file:

   • Filename should be {command-name}.md
   • Use frontmatter for advanced features (optional)
   • Write clear, concise prompt instructions

4. Add appropriate frontmatter fields (if needed):

   • allowed-tools: Restrict which tools Claude can use
   • argument-hint: Help text shown to users (e.g., [message], [PR number])
   • description: Brief description of the command
   • model: Specific model to use (e.g., claude-3-5-haiku-20241022)
   • disable-model-invocation: true: Prevent Claude from auto-invoking this command

5. Test the command:

   • Inform user they can use it immediately by typing /{command-name}
   • Show example usage with arguments if applicable

Frontmatter Template

---
allowed-tools: Tool1, Tool2, Bash(command:*)
argument-hint: [your hint here]
description: Brief description
model: claude-3-5-haiku-20241022
disable-model-invocation: false
---

Common Patterns

Simple Prompt Command

No frontmatter needed, just plain text:

Review this code for best practices and suggest improvements.

Command with Arguments

Use $ARGUMENTS for all args or $1, $2 for individual args:

---
argument-hint: [issue number]
description: Fix a specific GitHub issue
---

Fix issue #$ARGUMENTS following our coding standards and create a commit.

Command with Bash Execution

Use ! prefix for bash commands:

---
allowed-tools: Bash(git status:*), Bash(git diff:*), Bash(git commit:*)
description: Create a git commit
---

## Context

- Current status: !`git status`
- Current diff: !`git diff HEAD`
- Current branch: !`git branch --show-current`

## Task

Create a commit based on the above changes.

Command with File References

Use @ prefix for files:

---
argument-hint: [file path]
description: Document a specific file
---

Add comprehensive documentation to @$ARGUMENTS including:

- Purpose and functionality
- Parameters and return values
- Usage examples
- Edge cases

Command with Restricted Tools

Limit Claude to specific tools for safety:

---
allowed-tools: Read, Grep, Glob
description: Read-only code analysis
---

Analyze the codebase for security vulnerabilities without making any changes.

Examples

Example 1: Simple Review Command

File: .claude/commands/review.md

Review this code for:

- Code quality and readability
- Potential bugs
- Performance issues
- Security vulnerabilities

Usage: /review

Example 2: PR Review with Arguments

File: .claude/commands/pr-review.md

---
argument-hint: [PR number]
description: Review a GitHub pull request
---

Review PR #$ARGUMENTS and provide:

1. Summary of changes
2. Potential issues
3. Suggestions for improvement
4. Approval recommendation

Usage: /pr-review 123

Example 3: Complex Commit Command

File: .claude/commands/commit.md

---
allowed-tools: Bash(git add:*), Bash(git status:*), Bash(git commit:*), Bash(git diff:*)
argument-hint: [commit message]
description: Create a git commit with context
---

## Current State

- Branch: !`git branch --show-current`
- Status: !`git status`
- Staged changes: !`git diff --cached`

## Task

Create a commit with message: $ARGUMENTS

Follow conventional commits format and include detailed description.

Usage: /commit feat: add user authentication

Example 4: Test Generator

File: .claude/commands/test.md

---
argument-hint: [file path]
description: Generate tests for a file
---

Generate comprehensive unit tests for @$ARGUMENTS including:

- Happy path scenarios
- Edge cases
- Error handling
- Mocking external dependencies

Use the same testing framework already used in the project.

Usage: /test src/utils/helpers.ts

Command Organization

Using Subdirectories

Organize commands by category:

.claude/commands/
├── git/
│   ├── commit.md        → /commit (shows as "project:git")
│   └── pr-review.md     → /pr-review (shows as "project:git")
├── testing/
│   ├── unit-test.md     → /unit-test (shows as "project:testing")
│   └── e2e-test.md      → /e2e-test (shows as "project:testing")
└── docs/
    └── document.md      → /document (shows as "project:docs")

Best Practices

1. Use descriptive names: /review-pr is better than /rpr
2. Keep commands focused: One command = one clear purpose
3. Add argument hints: Help users understand what to pass
4. Use frontmatter sparingly: Only add it when you need the features
5. Test with different inputs: Verify arguments work as expected
6. Document complex commands: Add comments explaining what the command does
7. Use allowed-tools for safety: Restrict tools for read-only or limited-scope commands
8. Choose the right location:
   • Project commands (.claude/commands/): Team workflows, project-specific
   • Personal commands (~/.claude/commands/): Your personal preferences

Troubleshooting

Command Not Found

• Verify file exists in .claude/commands/ or ~/.claude/commands/
• Check filename matches command name (e.g., review.md for /review)
• May need to restart Claude Code to pick up new commands

Arguments Not Working

• Use $ARGUMENTS for all args or $1, $2, etc. for individual args
• Add argument-hint to frontmatter for better UX
• Test with and without arguments

Bash Commands Not Executing

• Use ! prefix: !git status``
• Add permission in allowed-tools: Bash(git status:*)
• Commands run in project directory

File References Not Working

• Use @ prefix: @src/file.ts
• Can use with arguments: @$1 or @$ARGUMENTS
• File must exist or Claude will ask for clarification

After Creating

After creating a command, remind the user:

1. The command is immediately available as /{command-name}
2. Show example usage with sample arguments
3. If it's a project command, they can commit it to git to share with team
4. They can edit the command file anytime to update it