Claude Code Slash Commands: The Complete Cheat Sheet (2026)

Slash commands are the fastest way to steer Claude Code, Anthropic's agentic command-line coding tool. A slash command is a keyword you type at the start of the prompt, beginning with a forward slash — like /clear or /help — that triggers a specific action instead of being sent to the model as an ordinary message. Some are built in and ship with the CLI; others are custom commands you define yourself as Markdown files. This cheat sheet covers both: a quick-reference table of the built-in commands, then everything you need to write and install your own.

Built-in slash commands cheat sheet

These commands are available in any Claude Code session out of the box. Type a forward slash in the prompt to see the list with autocomplete, or scan the table below. The exact set can grow between releases, so run /help to confirm what your version supports.

A few of these are worth committing to memory. /clear and /compact are your context-management workhorses: reach for /clear when you switch to an unrelated task, and /compact when a long session is filling the window but you still need its thread of reasoning. /init is usually the first thing to run in a new repository, because a good CLAUDE.md pays off in every later session.

Custom slash commands

Beyond the built-ins, you can define your own commands to replay a prompt you use often — generating a commit message, running a review pass, scaffolding a component, or writing tests for changed files. A custom slash command is nothing more than a Markdown file whose contents become the prompt when you invoke it.

Claude Code loads custom commands from two locations, and the scope decides where the command is available:

• Project commands live in .claude/commands/ at the root of your repository. Commit them and your whole team shares the same commands.
• Personal commands live in ~/.claude/commands/ in your home directory and follow you across every project on your machine.

The file name becomes the command name: a file called fix-tests.md is invoked by typing /fix-tests. Subdirectories act as namespaces, so .claude/commands/git/commit.md becomes /git:commit. No registration or restart is needed — Claude Code discovers the file the moment it exists.

Arguments and placeholders

Commands become far more useful when they accept input. Inside the Markdown body, the $ARGUMENTS placeholder is replaced by everything you type after the command name, so /fix-issue 123 high-priority passes 123 high-priority into the prompt. For finer control, use positional placeholders $1, $2, and so on to capture individual arguments separately — handy when the first token is an issue number and the second is a label.

An example command file

A command file can open with optional YAML frontmatter that configures how it runs, followed by the prompt itself. Save this as .claude/commands/review-diff.md:

---
description: Review the current git diff for bugs and style
argument-hint: [optional focus area]
allowed-tools: Bash(git diff:*), Read, Grep
---

Review the staged changes below for correctness, readability, and
security, focusing on $ARGUMENTS. Run git diff to see what changed,
then list findings grouped by severity with concrete fixes.

The description shows up in the command list, argument-hint documents expected input for autocomplete, and allowed-tools pre-approves specific tool calls so the command runs without extra prompts. You can also reference files with the @ prefix and run shell commands whose output is injected into the prompt using a leading !, which lets a command gather live context before Claude acts on it.

Installing ready-made commands

You do not have to write every command from scratch. ToolZip's Claude Code commands catalog collects working, categorized commands you can install with a single line. Each listing shows the command's contents and a copy-paste install command that uses the open-source claude-code-templates CLI:

npx claude-code-templates@latest --command="git/commit-message" --yes

Replace git/commit-message with the category/name path shown on any command page, run it from your project root, and the Markdown file drops straight into .claude/commands/ ready to use. The --yes flag skips the confirmation prompt for non-interactive setup. A fast way to learn the format is to install a command close to what you need, then open the generated file and edit the prompt to fit your workflow.

Tips for working with slash commands

• Type a forward slash to browse. Autocomplete lists both built-in and custom commands, so you rarely need to memorize names.
• Manage context deliberately. Use /clear between unrelated tasks and /compact when a long session is running low on room — both keep Claude sharp and responses fast.
• Write action-oriented descriptions. A clear description in a custom command's frontmatter makes it easy to find later and easy for teammates to understand.
• Scope tools with allowed-tools. Pre-approving just the tools a command needs keeps it fast without handing it broad permissions.
• Commit project commands to git. Storing them in .claude/commands/ gives your whole team the same shortcuts and versions them alongside the code.
• Namespace with folders. Group related commands into subdirectories so /git:commit and /test:unit stay organized as your collection grows.

Between the built-in commands in the table above and a handful of custom commands tuned to your project, slash commands turn repetitive instructions into single keystrokes. For a deeper walkthrough of the custom-command format, see the Claude Code commands guide, browse ready-made options in the commands catalog, or step back to the full Claude Code catalog to see how commands fit alongside agents, MCP servers, hooks, and skills.