← All guides

CommandsSetup

Claude Code Slash Commands: The Complete List (Built-In + Custom)

By Neo Zino - builder of ClockedCode·Updated September 2, 2026·16 min read

Every built-in Claude Code slash command by category, how to write your own (now merged with skills), the exact argument syntax, and why one might not show up.

Made with DispatchSEO

A Claude Code slash command is a /word you type at the start of a message to trigger a specific action instead of writing a prompt. Some are built in and run fixed logic straight from the CLI - /compact, /model, /config - while others are bundled skills that hand Claude a detailed set of instructions to follow, like /code-review or /debug. You can also write your own: save a file, and it becomes a working slash command the moment Claude Code notices it.

TL;DR: Claude Code ships 89 active built-in commands across nine rough categories (session control, memory, model/context, code review, workflow, permissions, MCP/plugins, design, and everything else) - the full tables are below. Writing your own no longer means learning a separate "custom commands" system: a file at .claude/commands/deploy.md and a skill at .claude/skills/deploy/SKILL.md create the identical /deploy command, arguments and all. Pass input with $ARGUMENTS, $1/$2, or a named $issue from the frontmatter, and if a command you wrote won't show up, claude plugin validate on its directory is the fastest way to find out why.

Every built-in slash command, by category

code.claude.com's own commands reference is one long alphabetical table with no section headers - the groupings below are this guide's own read of that table, done by counting every row while writing this guide on Claude Code 2.1.258. One listed command, /pr-comments, is marked removed as of v2.1.91 and isn't counted below; ask Claude to view PR comments directly instead.

Session & conversation control (13)

Memory & project setup (4)

Model, context & performance (8)

Code quality & review (11)

Workflow & automation (6)

Permissions, hooks & config (5)

MCP, plugins & agents (5)

Design & visualization (4)

Account, device & everything else (33)

That table (and the ones above it) is a direct count of code.claude.com's commands reference, not a copy of its wording - always check the official reference for the exact current text, since new commands land nearly every release.

How to create a custom slash command

Save a markdown file and it becomes a slash command - that's still true, and still the fastest path if you just want /your-name to exist. Create .claude/commands/your-name.md (project-only) or ~/.claude/commands/your-name.md (every project), add a description in YAML frontmatter, and write what you want Claude to do underneath:

---
description: Draft PR description bullets from the current diff and recent commits
---

## Diff
!`git diff main...HEAD`

## Your task
From the diff above, write 3-5 bullet points for a PR description.

That !`git diff main...HEAD` line runs before Claude ever sees the file - the command's actual output replaces the placeholder, so the instructions arrive with your real diff already inlined. Save the file and /your-name works immediately, no restart needed.

The fuller version of this - every frontmatter field, SKILL.md's supporting-files directory, invocation control, running a command as a subagent - lives in ClockedCode's Claude Code skills guide, since as of recent Claude Code releases that's the same system this section is describing. If you'd rather answer seven quick questions than write the frontmatter by hand, the free custom commands generator does the same job with a live preview.

Are custom commands and skills the same thing now?

Yes, and it's the detail that trips up anyone whose mental model is a year old: custom commands were folded into the skills system. A file at .claude/commands/deploy.md and a skill at .claude/skills/deploy/SKILL.md both create /deploy and run identically - your existing .claude/commands/ files keep working exactly as they are, with nothing to migrate. What a SKILL.md adds on top is optional: a directory for supporting reference files, frontmatter fields like disable-model-invocation to control who can trigger it, and the ability for Claude to load it automatically when your request matches its description - a .claude/commands/*.md file only ever runs when you type its name.

If a command and a skill share a name, the skill wins. And if you only take one thing from this section: there's no reason to keep writing new commands the old way. .claude/commands/ isn't deprecated, but every feature this guide covers past this point - named arguments, invocation control, running as a subagent - only exists on the SKILL.md side.

How arguments work in a custom slash command

Everything typed after the command name is available as $ARGUMENTS inside the command body. Running /fix-issue 123 against a command containing Fix GitHub issue $ARGUMENTS following our coding standards sends Claude that exact sentence with 123 substituted in.

Quote a multi-word value to pass it as one argument - /my-command "hello world" second makes $0 expand to hello world and $1 to second. An indexed placeholder with nothing at its position, like $2 when only one argument was given, is left in the text unchanged rather than turning into an empty string; a named placeholder from arguments: does the opposite and quietly becomes empty. Worth knowing before you ship a command that silently does the wrong thing when someone forgets an argument.

When a slash command is the wrong tool

• The trigger should be automatic, not typed. A command only runs when someone types its name. If you want Claude to reach for the instructions on its own whenever a request matches, that's the default behavior for a skill without disable-model-invocation set - see the skills guide for the invocation controls.
• It needs to fire on every tool call, not on request. A slash command is something a person or Claude chooses to run. Something that must happen every time regardless of what's decided - blocking a dangerous command, logging every edit - is a hook's job, not a command's.
• It's a fact true on every turn. A command's content only reaches Claude when it's invoked. A standing fact about the project belongs in CLAUDE.md, where it's loaded every time without anyone having to remember to type anything.
• It needs to call an external system. A command is prose Claude follows with the tools it already has - it can't add a new capability. Talking to a live API, a database, or a ticketing system is an MCP server's job.

Why a slash command isn't showing up or isn't working

Most of the time it's one of four things, roughly in order of how often each one turns out to be it:

1. Malformed frontmatter YAML. Claude Code still loads the file and /your-name still runs, but with empty metadata - so Claude has no description to automatically match your requests against. Run claude plugin validate .claude/commands (or .claude/skills) to find the exact parse error, or start with claude --debug to watch it load.
2. A name collision. Enterprise beats personal beats project; any of those beats a bundled command; a skill beats a same-named .claude/commands/*.md file. If /deploy isn't doing what you expect, something with the same name and higher priority is probably winning.
3. A brand-new top-level skills folder. Live editing an existing .claude/skills/ or .claude/commands/ directory picks up changes immediately, mid-session. Creating the first one in a project that didn't have it yet needs a session restart before Claude Code starts watching it.
4. It's plainly platform- or plan-gated, and you haven't noticed. A handful of built-in commands only show up under specific conditions - /desktop only on macOS or x64 Windows with a subscription, /upgrade only when an update exists and never on Enterprise, /heapdump hidden from the menu on purpose. Typing the full name still works for a hidden command even though it won't appear as you type /.

FAQ

How many slash commands does Claude Code have?

89 active built-in commands as of Claude Code 2.1.258, counted directly from the official commands reference (a 90th, /pr-comments, is listed but marked removed). That number grows with nearly every release, so treat it as a snapshot rather than a permanent count.

Are custom slash commands the same as skills now?

Functionally yes. A file at .claude/commands/deploy.md and a skill at .claude/skills/deploy/SKILL.md both create /deploy and behave the same way - custom commands were folded into the skills system. The old .claude/commands/ location still works; skills just add optional extras like a directory for supporting files and finer invocation control.

How do I pass arguments to a custom slash command?

Everything typed after the command name is available as $ARGUMENTS in the command body. To use individual words, $1 and $2 (or $ARGUMENTS[0] and $ARGUMENTS[1]) pull out the first and second argument by position, and a named arguments: list in the frontmatter lets you reference them as $issue or $branch instead of by number.

Why isn't my custom slash command showing up?

The most common cause is malformed frontmatter YAML: Claude Code still loads the file, but with empty metadata, so the command works when typed by name but Claude has no description to match against for automatic use. Run claude plugin validate on the .claude/commands or .claude/skills directory to find the exact parse error, or start Claude Code with --debug.

Can I create my own slash commands in Claude Code?

Yes. Save a markdown file to .claude/commands/your-name.md, or a SKILL.md to .claude/skills/your-name/SKILL.md for the fuller feature set, and it becomes /your-name the moment the file exists. ClockedCode's free custom commands generator asks 7 quick questions and hands back a working command file with valid frontmatter.

Ninety commands is too many to hold in your head, which is exactly why /help and typing / to filter exist - this list is for the moment you need to know a command is real before you go looking for it, or that the thing you want (self-paced background loops, a subagent that reports back without detaching, a hidden memory diagnostic) already shipped under a name you hadn't tried yet. ClockedCode's curated setup and free generators exist for the same reason: less time rediscovering what Claude Code already does, more time using it.