DEV Community

brian austin
brian austin

Posted on

Claude Code .claude/commands: how to build your own slash commands library

#claudecode #ai #productivity #devtools

Claude Code .claude/commands: how to build your own slash commands library

Claude Code ships with built-in slash commands like /clear, /compact, /review, and /help. But the most useful ones are the ones you build yourself.

Here's how to create a personal commands library that works across every project.

Where custom commands live 

your-project/
├── .claude/
│   └── commands/
│       ├── deploy.md
│       ├── review.md
│       ├── debug.md
│       └── explain.md
Enter fullscreen mode Exit fullscreen mode

Each .md file becomes a slash command. deploy.md → /project:deploy.

You can also put them in ~/.claude/commands/ to make them available globally across all projects.

Your first custom command

Create .claude/commands/review.md:

Review the code I just wrote:

1. Check for security issues (SQL injection, XSS, auth bypasses)
2. Check for performance issues (N+1 queries, unnecessary loops)
3. Check error handling — are edge cases covered?
4. Suggest one specific refactor that would improve readability

Format as a numbered list. Be direct. Don't compliment the code.
Enter fullscreen mode Exit fullscreen mode

Now type /project:review in any Claude Code session and it runs this exact prompt.

Useful commands to steal

/project:explain — explain a file for a new developer 

Explain $ARGUMENTS as if I'm a new developer joining this project:

- What does this file/function do?
- What are the inputs and outputs?
- What could go wrong?
- What would I need to understand before modifying it?

Keep it under 200 words. Use plain language.
Enter fullscreen mode Exit fullscreen mode

Usage: /project:explain src/auth/middleware.js

The $ARGUMENTS variable captures whatever you type after the command name.

/project:debug — structured debugging session 

I have a bug. Let's debug it systematically.

Bug description: $ARGUMENTS

1. First, list 3 most likely root causes
2. For each, describe how to verify it in 1-2 steps
3. Once we identify the cause, propose a minimal fix
4. Flag any tests that need updating

Do not write code until we've identified the root cause.
Enter fullscreen mode Exit fullscreen mode

/project:commit — generate a commit message 

Generate a conventional commit message for the changes in the current git diff.

Format:
Enter fullscreen mode Exit fullscreen mode

(): 



Types: feat, fix, docs, style, refactor, test, chore
Keep the description under 72 characters.
Focus on WHY, not WHAT.
Enter fullscreen mode Exit fullscreen mode

/project:deploy — pre-deploy checklist 

Run my pre-deploy checklist:

1. Check for any TODO/FIXME/HACK comments in recently changed files
2. Verify environment variables are documented in README
3. Check if any new dependencies need security audit (`npm audit`)
4. Confirm no hardcoded secrets (scan for common patterns)
5. List any database migrations needed

If anything needs attention, stop and flag it. Don't proceed.
Enter fullscreen mode Exit fullscreen mode

/project:test-plan — generate test cases 

For the code I just wrote ($ARGUMENTS), generate a test plan:

1. Happy path tests (3 minimum)
2. Edge cases (boundary values, empty inputs, null)
3. Error conditions (what should fail gracefully)
4. Integration points (what other systems does this touch?)

For each test, write a one-line description. I'll implement them.
Enter fullscreen mode Exit fullscreen mode

Global commands for your home directory

Put these in ~/.claude/commands/ so they work everywhere:

~/.claude/commands/standup.md 

Generate my standup update based on recent git commits:
Enter fullscreen mode Exit fullscreen mode


bash
git log --oneline --since=yesterday --author=$(git config user.email)


Format:
- Yesterday: [what I completed]
- Today: [what I'm working on based on current branch/uncommitted changes]
- Blockers: [anything in TODO comments or failing tests]

Keep it under 5 bullets total.
Enter fullscreen mode Exit fullscreen mode


markdown

~/.claude/commands/pr-description.md 

Write a pull request description for these changes.

Include:
- Summary (2-3 sentences, non-technical)
- What changed and why
- How to test it
- Screenshots needed? (yes/no)

Use the diff and commit messages as context.
Enter fullscreen mode Exit fullscreen mode

The $ARGUMENTS trick

Any text after your command name becomes $ARGUMENTS:

/project:explain auth/middleware.js
Enter fullscreen mode Exit fullscreen mode

Inside your command file, $ARGUMENTS = auth/middleware.js.

Use this to make commands flexible without making them verbose.

Commands that link to SimplyLouie's API

If you're running Claude Code against a budget-friendly API backend, add the endpoint to your commands:

For this task, use ANTHROPIC_BASE_URL=https://simplylouie.com/api — it removes the rate limit interruptions that break long sessions.

Task: $ARGUMENTS
Enter fullscreen mode Exit fullscreen mode

This is useful when you're building automation scripts that call Claude directly — at $2/month via SimplyLouie, you get the full API without the per-session rate walls.

Organizing a large command library

Once you have 10+ commands, organization matters:

.claude/commands/
├── git/
│   ├── commit.md
│   └── pr-description.md
├── code/
│   ├── review.md
│   ├── explain.md
│   └── debug.md
└── ops/
    ├── deploy.md
    └── standup.md
Enter fullscreen mode Exit fullscreen mode

Subfolder commands become /project:git:commit, /project:code:review, etc.

The meta-command: /project:new-command

My favorite trick — a command that generates new commands:

I want to create a new Claude Code slash command.

Task I want to automate: $ARGUMENTS

1. Write a command prompt that handles this task well
2. Suggest a filename (all lowercase, hyphens for spaces)
3. Include $ARGUMENTS if the task needs variable input
4. Add it to `.claude/commands/`
Enter fullscreen mode Exit fullscreen mode

Now you can type /project:new-command generate test fixtures for any model and Claude will write the command file for you.

The actual benefit

Custom commands solve a specific problem: you've found the perfect prompt for a task, but you can't remember it next time.

Instead of:

  • Searching your chat history for that one good prompt
  • Rewriting it slightly wrong every time
  • Copying from a notes app

You have /project:review and it's identical every time.

The library builds up naturally. Every time you write a prompt that works well, extract it into a command.

Running Claude Code sessions? SimplyLouie provides Claude API access at $2/month — same models, no rate limit interruptions during long coding sessions.

Top comments (0)