writing-slash-commands

Slash Commands Guide

Slash commands are reusable prompt templates that automate recurring tasks in Claude Code. Define them once, invoke them anytime with /command-name [args].

Quick Start

Create a Command

Project-scoped (shared with team):

mkdir -p .claude/commands
echo "Analyze this code for performance issues:" > .claude/commands/optimize.md

Personal (across all projects):

mkdir -p ~/.claude/commands
echo "Review this code for security vulnerabilities:" > ~/.claude/commands/security-review.md

Invoke with /optimize or /security-review.

---

Custom Commands Architecture

Scope	Location	Scope	Shareable
Project	.claude/commands/	Repository-specific	Yes (team)
Personal	~/.claude/commands/	All projects	No

Namespacing

Organize commands in subdirectories (no effect on invocation):

.claude/commands/
├── frontend/
│   └── component.md      # Invokes as `/component` (shows "project:frontend")
├── backend/
│   └── api.md           # Invokes as `/api` (shows "project:backend")
└── security-review.md   # Invokes as `/security-review`

Note: User-level and project-level commands with the same name conflict; only one is available.

---

Arguments & Placeholders

Capture All Arguments with $ARGUMENTS

Use $ARGUMENTS when you need all arguments as a single string:

---
description: Fix issue with optional priority
argument-hint: [issue-number] [optional-priority]
---

Fix issue #$ARGUMENTS following our coding standards.

Usage: /fix-issue 123 → $ARGUMENTS = "123" Usage: /fix-issue 123 high-priority → $ARGUMENTS = "123 high-priority"

Access Individual Arguments with $1, $2, ...

Use positional parameters for structured commands:

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

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

Usage: /review-pr 456 high alice → $1="456", $2="high", $3="alice"

When to use positional:

• Arguments have distinct roles (ID, priority, owner)
• Need defaults: ${3:-unassigned}
• Reference arguments separately throughout command

---

Frontmatter Configuration

All options are optional; commands work without frontmatter.

Field	Purpose	Example
description	Shown in /help (required for SlashCommand tool)	"Fix security issues"
argument-hint	Argument syntax hint for autocomplete	"[issue] [priority]"
allowed-tools	Tools this command can invoke	Bash(git add:*), Bash(git status:*)
model	Override default model for this command	"claude-3-5-haiku-20241022"
disable-model-invocation	Prevent SlashCommand tool from triggering	true

Example with Full Frontmatter

---
description: Create a git commit with staged changes
argument-hint: [message]
allowed-tools: Bash(git add:*), Bash(git status:*), Bash(git commit:*)
model: claude-3-5-haiku-20241022
---

Create a git commit with message: $ARGUMENTS

---

Bash Execution & File References

Run Bash Commands with !

Prefix inline commands with ! to execute before the command runs. Requires allowed-tools with Bash:

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

## Context

- Current status: !`git status`
- Staged changes: !`git diff --cached`
- Recent commits: !`git log --oneline -5`

## Your Task

Create a single commit summarizing the changes.

Output from bash commands is included in the prompt context.

Reference Files with @

Include file contents in commands:

Review the implementation in @src/utils/helpers.js

Compare @src/old-version.js with @src/new-version.js

Use standard file references (e.g., @docs/, @src/).

---

Pattern Examples

Example 1: Priority-Based Issue Fix

---
description: Fix issue with priority level
argument-hint: [issue-number] [priority-level]
---

Fix GitHub issue #$1 with priority "$2".

Steps:
1. Understand the issue context
2. Write minimal, focused fix
3. Consider edge cases
4. Ensure tests pass

Usage: /fix-issue 42 high

---

Example 2: Bash-Powered Code Review

---
allowed-tools: Bash(git diff:*), Bash(git log:*)
description: Review recent commits for code quality
argument-hint: "[number-of-commits]"
---

## Context

Recent changes:
!`git log --oneline -${1:-5}`

Full diff:
!`git diff HEAD~${1:-5}...HEAD`

## Your Task

Provide a code quality review focusing on readability, performance, and best practices.

Usage: /review-commits 3 → Reviews last 3 commits

---

Example 3: Multi-Argument Configuration Command

---
description: Set up feature flags for testing
argument-hint: feature [enable|disable] [environment]
---

Configure feature "$1" to be $2 in the $3 environment.

Verify:
- Feature flag exists in @src/config/features.ts
- Environment is valid (dev, staging, production)
- Changes are tested before deployment

Usage: /feature dark-mode enable staging

---

SlashCommand Tool Integration

The SlashCommand tool allows Claude to invoke your custom commands programmatically.

Enable Auto-Invocation

Add to CLAUDE.md or project instructions:

When appropriate, use /optimize to analyze code performance.
When fixing bugs, use /fix-issue with the issue number.

Requirements for Tool Access

1. Command must have description frontmatter
2. Command must NOT have disable-model-invocation: true
3. User must allow SlashCommand tool in permissions

Disable Specific Commands

Prevent Claude from auto-invoking a command:

---
disable-model-invocation: true
description: Manual review only
---

Your command content...

Permission Rules

Fine-grained control via /permissions:

SlashCommand:/commit         # Exact match: /commit only
SlashCommand:/review-pr:*    # Prefix match: /review-pr with any args

Deny SlashCommand entirely to disable all auto-invocation.

---

Best Practices

✅ Do:

• Use descriptive names matching functionality (e.g., /optimize, /security-review)
• Include description for discoverability via /help and SlashCommand tool
• Add argument-hint for clear usage patterns
• Keep commands focused on a single responsibility
• Use bash execution for context that changes (git status, timestamps)

❌ Don't:

• Embed static content that belongs in code (use file references instead)
• Create commands without descriptions (breaks tool integration)
• Overload with too many positional arguments (>3 becomes hard to remember)
• Assume tools are available without declaring allowed-tools

---

Built-in Slash Commands (Reference)

Essential commands you get for free:

Command	Purpose
/help	List all commands (built-in + custom)
/config	Open Settings interface
/status	Show version, model, account
/cost	Token usage statistics
/model	Switch AI model
/memory	Edit CLAUDE.md
/rewind	Rewind conversation or code
/clear	Clear history
/agents	Manage custom AI subagents
/mcp	Manage MCP server connections

---

Workflow Integration

In CLAUDE.md or project instructions:

## Custom Commands

Use these commands to accelerate development:

- `/optimize [file]` — Analyze code performance
- `/security-review [file]` — Check for vulnerabilities
- `/commit [message]` — Create atomic commits
- `/test [suite]` — Run tests with focus

In conversations:

> I'll use /optimize to check this function for performance issues.

Claude recognizes the slash and may auto-invoke if SlashCommand tool is enabled.

---

See Also

• MCP Slash Commands: Commands exposed by MCP servers (pattern: /mcp__server__prompt)
• Plugin Commands: Commands from installed plugins (pattern: /plugin-name:command or just /command)
• Interactive Mode: Keyboard shortcuts and input modes
• Permissions: Fine-grained tool and command access control