Slash Commands

Overview

Slash commands control Claude's behavior during interactive sessions. They're organized into built-in commands, custom commands defined by users, plugin commands, and MCP-exposed prompts.

Built-in Commands

Key built-in commands include:

• /help - Get usage help
• /clear - Clear conversation history
• /model - Select or change the AI model
• /cost - Show token usage statistics
• /memory - Edit CLAUDE.md memory files
• /sandbox - Enable sandboxed bash execution
• /rewind - Rewind conversation and/or code
• /mcp - Manage MCP server connections

Custom Slash Commands

File Organization

Custom commands are Markdown files stored in two locations:

• Project commands: .claude/commands/ (shared with team via git)
• Personal commands: ~/.claude/commands/ (user-level only)

Basic Syntax

/<command-name> [arguments]

Arguments

Commands support two argument styles:

Capture all arguments:

Fix issue #$ARGUMENTS following coding standards

Individual positional arguments:

Review PR #$1 with priority $2 and assign to $3

Features

Namespacing: Organize commands in subdirectories. A file at .claude/commands/frontend/component.md creates /component with description "(project:frontend)".

Bash execution: Prefix commands with ! to execute bash and include output:

---
allowed-tools: Bash(git status:*), Bash(git add:*)
---
Current git status: ![git status]

Note: Replace [git status] with the actual command in backticks when creating your slash command.

File references: Use @ prefix to include file contents in commands:

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

Frontmatter metadata:

---
description: Brief command description
allowed-tools: Tool specifications
argument-hint: [expected-args]
model: specific-model-id
---

Plugin Commands

Plugins distribute custom slash commands automatically. They use the format /plugin-name:command-name when disambiguation is needed, support all standard features (arguments, bash execution, file references), and appear in /help when installed.

MCP Slash Commands

MCP servers expose prompts as commands with the pattern:

/mcp__<server-name>__<prompt-name> [arguments]

These are dynamically discovered from connected MCP servers and automatically available when the server is active.

SlashCommand Tool

The SlashCommand tool allows Claude to invoke custom commands programmatically during conversations. To encourage usage, reference commands in your instructions with their slash prefix (e.g., "Run /write-unit-test").

Limitations: Only supports user-defined custom commands with a description field populated in frontmatter. Built-in commands like /compact cannot be invoked via the tool.

Management:

• Disable all: /permissions then add SlashCommand to deny rules
• Disable specific commands: Add disable-model-invocation: true to frontmatter
• Permission syntax: SlashCommand:/commit (exact match) or SlashCommand:/review-pr:* (prefix match)

Skills vs Slash Commands

Use slash commands for: Simple, frequently-used prompts that fit in one file.

Use Skills for: Complex capabilities requiring multiple files, scripts, or organizational structure.

Key difference: Commands require explicit invocation; Skills are discovered automatically based on context.