January 12, 202611 MIN READ

Custom Slash Commands in Claude Code: Build Your Own

By Learnia Team

Custom Slash Commands in Claude Code: Build Your Own

This article is written in English. Our training modules are available in French.

Built-in slash commands cover common operations, but every team has unique workflows. Custom commands let you package complex prompts into simple triggers—turn multi-step processes into single commands that anyone can run.

Why Custom Commands?

Without custom commands, you repeat yourself:

> Review this code following our style guide. Check for:
  - ESLint compliance
  - TypeScript strict mode compatibility
  - Our naming conventions from CONTRIBUTING.md
  - Performance patterns we use
  - Security vulnerabilities
  Run the tests after suggesting fixes.

With custom commands:

> /code-review

Same result, zero typing. And your entire team uses the exact same process.

Command File Structure

Custom commands are markdown files in

.claude/commands/

→project/.claude/commands/deploy.md
→project/.claude/commands/code-review.md
→project/.claude/commands/generate-docs.md

Basic Command

<!-- .claude/commands/hello.md -->
---
description: Say hello
---

Say hello to the user and introduce yourself.

Usage:

/hello

Command with Parameters

<!-- .claude/commands/create-component.md -->
---
description: Create a React component
---

Create a new React component named $COMPONENT_NAME in the src/components folder.

Include:
- TypeScript types
- CSS module for styling
- Unit tests in __tests__ folder
- Storybook story

Usage:

/create-component Button

The

$COMPONENT_NAME

becomes "Button".

Frontmatter Options

The YAML frontmatter configures command behavior:

---
description: Short description shown in /help
allowed-tools: Tool1, Tool2(pattern)
---

description (required)

Appears in command listings:

> /help

Custom Commands:
  /deploy         Deploy to staging environment
  /code-review    Review code with team standards
  /generate-docs  Generate API documentation

allowed-tools (optional)

Restricts which tools the command can use:

---
description: Safe documentation generator
allowed-tools: Read, Write(docs/**)
---

Generate documentation for the current project.

This command can only read files and write to the docs folder.

Tool patterns:

Pattern	Meaning
`Read`	All read operations
`Write(docs/**)`	Write only to docs folder
`Bash(npm:*)`	Only npm commands
`Edit(src/**)`	Edit only src files

Creating Your First Command

Step 1: Create the Commands Folder

mkdir -p .claude/commands

Step 2: Write the Command File

<!-- .claude/commands/review.md -->
---
description: Review code following team standards
allowed-tools: Read, Bash(npm:test), Bash(npm:lint)
---

Review the current changes following our team standards:

## Code Quality
- Check ESLint compliance
- Verify TypeScript types are correct
- Look for unused imports/variables

## Style Guide
- Follow naming conventions in CONTRIBUTING.md
- Ensure consistent formatting

## Security
- Check for hardcoded secrets
- Look for SQL injection vulnerabilities
- Verify input validation

## Performance
- Identify N+1 queries
- Check for unnecessary re-renders
- Look for memory leaks

After review, run `npm test` and `npm run lint` to verify changes.

Step 3: Use the Command

> /review

Claude follows your template exactly.

Parameter Substitution

Commands support dynamic values using

$PARAMETER_NAME

Single Parameter

<!-- .claude/commands/test.md -->
---
description: Run tests for a specific module
---

Run tests for the $MODULE module:

1. Find all test files in src/$MODULE
2. Run `npm test -- --testPathPattern=$MODULE`
3. Summarize results

Usage:

> /test auth
> /test payments

Multiple Parameters

<!-- .claude/commands/migrate.md -->
---
description: Create database migration
---

Create a new database migration:

- Name: $MIGRATION_NAME
- Type: $TYPE (create_table, alter_table, add_index)

Generate the migration file and corresponding rollback.

Usage:

> /migrate add_user_avatar alter_table

Default Values

<!-- .claude/commands/build.md -->
---
description: Build for environment
---

Build the project for $ENV environment.

$ENV defaults to "development" if not specified.

Context Variables

Commands can reference contextual information:

Variable	Description
`$CURRENT_FILE`	Currently active file
`$SELECTION`	Selected text in IDE
`$ARGUMENTS`	All arguments as string

<!-- .claude/commands/explain.md -->
---
description: Explain selected code
---

Explain this code in detail:

$SELECTION

Cover:
- What it does
- How it works
- Why it's written this way

Real-World Command Examples

Code Review Command

<!-- .claude/commands/code-review.md -->
---
description: Comprehensive code review
allowed-tools: Read, Bash(git:diff*), Bash(npm:test), Bash(npm:lint)
---

Perform a thorough code review of recent changes.

## 1. Gather Changes
Run `git diff HEAD~1` to see recent changes.

## 2. Review Checklist

### Logic & Correctness
- [ ] Does the code do what it's supposed to?
- [ ] Are edge cases handled?
- [ ] Is error handling appropriate?

### Code Quality
- [ ] Is the code readable and well-organized?
- [ ] Are functions/methods appropriately sized?
- [ ] Is there unnecessary duplication?

### Testing
- [ ] Are new features tested?
- [ ] Do existing tests still pass?

### Security
- [ ] Is user input validated?
- [ ] Are there any exposed secrets?
- [ ] Are there SQL injection risks?

## 3. Run Verification
Run `npm test` and `npm run lint`.

## 4. Summary
Provide a summary with:
- Overall assessment (Approve/Request Changes)
- List of issues found
- Suggestions for improvement

Deploy Command

<!-- .claude/commands/deploy.md -->
---
description: Deploy to staging
allowed-tools: Bash(npm:*), Bash(git:*), Bash(docker:*)
---

Deploy the current branch to staging:

## Pre-flight Checks
1. Ensure all tests pass: `npm test`
2. Check for uncommitted changes: `git status`
3. Verify on correct branch

## Build
1. Run production build: `npm run build`
2. Build Docker image: `docker build -t app:staging .`

## Deploy
1. Push image: `docker push registry/app:staging`
2. Update staging: `kubectl rollout restart deployment/app -n staging`

## Verify
1. Wait for rollout: `kubectl rollout status deployment/app -n staging`
2. Check health endpoint
3. Run smoke tests

## Report
Summarize:
- Commit deployed
- Time taken
- Any issues encountered

API Endpoint Generator

<!-- .claude/commands/create-endpoint.md -->
---
description: Generate a new API endpoint
allowed-tools: Read, Write, Edit
---

Create a new REST API endpoint:

## Endpoint Details
- Resource: $RESOURCE
- Method: $METHOD (GET, POST, PUT, DELETE)

## Generate Files

### Route Handler
Create `src/routes/$RESOURCE.ts`:
- Express/Fastify route handler
- Input validation with Zod
- Error handling middleware

### Service Layer
Create `src/services/$RESOURCE.service.ts`:
- Business logic
- Database operations
- Proper error types

### Tests
Create `src/__tests__/$RESOURCE.test.ts`:
- Unit tests for service
- Integration tests for route
- Mock database calls

### Types
Add types to `src/types/$RESOURCE.ts`:
- Request/response interfaces
- Domain models

## Documentation
Update `docs/api.md` with new endpoint documentation.

Usage:

> /create-endpoint users POST
> /create-endpoint products GET

Documentation Generator

<!-- .claude/commands/document.md -->
---
description: Generate documentation for a file
allowed-tools: Read, Write(docs/**)
---

Generate comprehensive documentation for $FILE:

## Analysis
1. Read the file and understand its purpose
2. Identify all exports (functions, classes, types)
3. Trace dependencies

## Documentation Output
Create `docs/$FILE.md` with:

### Overview
- File purpose
- When to use it
- Dependencies

### API Reference
For each export:
- Description
- Parameters with types
- Return value
- Example usage

### Examples
Provide 2-3 practical examples showing common use cases.

### Related
Link to related documentation files.

User-Level Commands

Personal commands go in

~/.claude/commands/

→~/.claude/commands/my-workflow.md
→~/.claude/commands/quick-fix.md

User commands are available in all projects.

Personal Productivity Command

<!-- ~/.claude/commands/standup.md -->
---
description: Generate standup update
allowed-tools: Read, Bash(git:log*)
---

Generate my daily standup update:

## Yesterday
Run `git log --oneline --since="yesterday" --author="<my-email>"` and summarize what I worked on.

## Today
Based on:
- Open TODOs in the codebase
- Recent branch names
- Uncommitted changes

Suggest what I should focus on today.

## Blockers
Check for:
- Failing tests
- Merge conflicts
- Pending reviews

Format as:
---
**Yesterday:** <summary>
**Today:** <plan>
**Blockers:** <issues or "None">
---

Command Organization

By Feature

Organize commands in subfolders:

→
auth/ -
```
add-oauth.md
```
,
```
update-session.md
```
→
api/ -
```
create-endpoint.md
```
,
```
document-api.md
```
→
testing/ -
```
unit-test.md
```
,
```
e2e-test.md
```

Subfolders become command prefixes:

> /auth/add-oauth
> /api/create-endpoint

By Role

Organize by team role:

→
dev/ -
```
review.md
```
,
```
refactor.md
```
→
qa/ -
```
test-plan.md
```
,
```
bug-report.md
```
→
devops/ -
```
deploy.md
```
,
```
rollback.md
```

Team Command Sharing

Version Control

Commit

.claude/commands/

to your repository:

git add .claude/commands/
git commit -m "Add team Claude Code commands"

Everyone gets the same commands automatically.

Command Standards

Create a README for your commands:

<!-- .claude/commands/README.md -->
# Team Commands

## Development
- `/review` - Code review with team standards
- `/create-component` - Generate React component

## DevOps
- `/deploy staging` - Deploy to staging
- `/deploy production` - Deploy to production (requires approval)

## Testing
- `/test-unit` - Run unit tests
- `/test-e2e` - Run end-to-end tests

## Adding Commands
1. Create a markdown file in `.claude/commands/`
2. Follow the template in `_template.md`
3. Submit a PR for team review

Advanced Patterns

Command Chaining

Reference other commands in your prompts:

<!-- .claude/commands/release.md -->
---
description: Full release process
---

Perform a complete release:

1. First, run the code review process (like /review)
2. Update version in package.json
3. Generate changelog from commits
4. Build for production (like /build production)
5. Create git tag
6. Deploy (like /deploy production)

Conditional Logic

Use prompts to guide conditional behavior:

<!-- .claude/commands/fix.md -->
---
description: Fix issues based on type
---

Fix the following issue: $ISSUE_TYPE

If $ISSUE_TYPE is "lint":
  - Run npm run lint:fix
  - Review and commit changes

If $ISSUE_TYPE is "test":
  - Run failing tests
  - Fix the failures
  - Verify all tests pass

If $ISSUE_TYPE is "type":
  - Run TypeScript compiler
  - Fix type errors
  - Ensure strict mode compliance

Interactive Commands

Commands can ask for more input:

<!-- .claude/commands/refactor.md -->
---
description: Interactive refactoring session
---

Start an interactive refactoring session.

First, ask the user:
1. What file or component to refactor?
2. What's the goal of the refactoring?
3. Are there any constraints?

Then:
1. Analyze the current code
2. Propose a refactoring plan
3. Wait for approval
4. Implement changes incrementally
5. Run tests after each change

Debugging Commands

Test Your Command

Run with verbose output:

claude --verbose
> /my-command

Common Issues

Command not found:

→
Check file is in
```
.claude/commands/
```
→
Verify filename ends in
```
.md
```
→
Ensure frontmatter has
```
description
```

Parameters not substituting:

→
Use exact format:
```
$PARAMETER_NAME
```
→Check for typos in variable names
→Ensure arguments are provided in order

Permission errors:

→
Review
```
allowed-tools
```
restrictions
→Check project permission settings
→
Use
```
/permissions
```
to debug

Integration with Other Features

Commands + Hooks

Trigger hooks from commands:

<!-- .claude/commands/safe-deploy.md -->
---
description: Deploy with safety hooks
allowed-tools: Bash(npm:*), Bash(git:*), Bash(docker:*)
---

Deploy to production with all safety checks.

This command relies on our PreToolUse hooks to:
- Validate deployment approvals
- Check for required tags
- Notify team via Slack

See Claude Code Hooks: Automate Your Development Workflow.

Commands + MCP

Use MCP tools in commands:

<!-- .claude/commands/sync-issues.md -->
---
description: Sync GitHub issues
allowed-tools: mcp__github__*, Read, Write
---

Sync GitHub issues with local tracking:

1. Use GitHub MCP to fetch open issues
2. Update local TODO.md with issue list
3. Mark completed issues as done

See Model Context Protocol (MCP) for Claude Code: Complete Guide.

Commands + Sub-Agents

Spawn specialized agents:

<!-- .claude/commands/deep-analysis.md -->
---
description: Deep codebase analysis
---

Perform a deep analysis of the codebase:

Use the Explore sub-agent to:
1. Map all modules and dependencies
2. Identify architectural patterns
3. Find potential issues

Then synthesize findings into a report.

See Claude Code Sub-Agents: Orchestrating Complex Tasks.

Key Takeaways

→
Commands are markdown files: Simple to create, version, and share.
→
Use parameters for flexibility:
```
$PARAM_NAME
```
makes commands reusable.
→
Restrict tools for safety:
```
allowed-tools
```
limits what commands can do.
→
Commit team commands: Share via version control for consistency.
→
Combine with other features: Commands integrate with hooks, MCP, and sub-agents.

Build Effective Prompts for Your Commands

Custom commands are only as good as the prompts inside them. Learn to write prompts that get consistent results.

In our Module 3 — Prompt Engineering Basics, you'll learn:

→Core principles of effective prompting
→Structure and clarity techniques
→Iterative prompt refinement
→Common patterns and anti-patterns

→ Explore Module 3: Prompt Engineering Basics

GO DEEPER

Module 3 — Chain-of-Thought & Reasoning

Master advanced reasoning techniques and Self-Consistency methods.

Explorer le Module