---

name: meta:command-creator description: Create custom slash commands for Claude Code with proper markdown structure and YAML frontmatter. Use when creating commands, generating slash commands, making workflow automations.

Slash Command Creator

Quick Start

Step 1: Create the command

Copy# .claude/commands/my-command.md

---

description: Brief description of what this command does.

---

## Steps

1. First action
2. Second action
3. Final action

Invoke with /my-command in Claude Code (shows "(project)" in /help).

Step 2: Validate (required)

Always run validation after creating or modifying a command:

Copyuv run .claude/skills/meta-command-creator/scripts/validate-command.py .claude/commands/my-command.md

Fix any errors before committing.

Command Locations

Priority

When project and user commands share the same name, project wins and user command is silently ignored.

Tip: Slash command autocomplete works anywhere in your input, not just at the beginning.

YAML Frontmatter (Official Fields Only)

Copy---
description: Create a conventional git commit with proper formatting.
allowed-tools: Bash(git add:*), Bash(git commit:*), Bash(git status:*)
argument-hint: [commit message]
---

Invalid Fields (Do Not Use)

• name - Inferred from filename
• category - Not an official field
• tags - Not an official field

Special Syntax

Bash Execution (! prefix)

Execute bash commands before the slash command runs. Output is included in context:

Copy---
allowed-tools: Bash(git:*)
description: Create a git commit
---

## Context

- Current status: !`git status`
- Current diff: !`git diff HEAD`
- Current branch: !`git branch --show-current`
- Recent commits: !`git log --oneline -10`

## Your task

Based on the above changes, create a single git commit.

Important: You MUST include allowed-tools with Bash when using ! prefix.

File References (@ prefix)

Include file contents in commands:

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

Compare @src/old-version.js with @src/new-version.js

Extended Thinking

Include extended thinking keywords to trigger deeper reasoning:

CopyThink carefully about the architecture implications before making changes.

Command Templates

See reference.md for complete templates:

• Basic Command - Simple steps with reference section
• Command with Arguments - Using $ARGUMENTS and $1, $2
• Guardrailed Command - Safety constraints for state-changing operations
• Multi-Phase Command - Complex workflows with distinct phases
• Git Commit Command - With bash execution context

$ARGUMENTS Patterns

See reference.md for detailed patterns.

Organizing Commands

Use subdirectories for categorization:

Copy.claude/commands/
├── git/commit.md      # Shows "(project:git)"
├── dev/feature.md     # Shows "(project:dev)"
└── docs/update.md     # Shows "(project:docs)"

SlashCommand Tool

Claude can execute custom slash commands programmatically using the SlashCommand tool.

Enabling Automatic Invocation

Reference commands by name in your prompts or CLAUDE.md:

CopyRun /write-unit-test when you are about to start writing tests.

Requirements

• Command MUST have description frontmatter (used for context)
• Built-in commands (/compact, /init, etc.) are NOT supported
• Only custom user-defined commands work

Disabling

Copy# Disable all SlashCommand tool invocations
/permissions
# Add to deny rules: SlashCommand

# Disable specific command only

Add to command frontmatter:

Copydisable-model-invocation: true

Permission Rules

• Exact match: SlashCommand:/commit (only /commit, no args)
• Prefix match: SlashCommand:/review-pr:* (allows any args)

Character Budget

Default: 15,000 characters for command descriptions. Set custom limit:

CopySLASH_COMMAND_TOOL_CHAR_BUDGET=20000

When exceeded, shows "M of N commands" warning in /context.

Skills vs Commands

When to Use Each

Use slash commands when:

• You invoke the same prompt repeatedly
• The prompt fits in a single file
• You want explicit control over when it runs

Use Skills when:

• Claude should discover the capability automatically
• Multiple files or scripts are needed
• Complex workflows with validation steps
• Team needs standardized, detailed guidance

Common Mistakes

Validation

Run the validator to check command structure:

Copyuv run .claude/skills/meta-command-creator/scripts/validate-command.py .claude/commands/my-command.md

Checklist

• File location: .claude/commands/<name>.md
• YAML frontmatter has description
• Steps are numbered and actionable
• $ARGUMENTS handled correctly (if used)
• allowed-tools included if using ! bash execution
• All code blocks have language specifiers
• Guardrails section for state-changing commands
• Output format specified

Built-in Commands Reference

Key built-in commands (cannot be overridden):

Full list: Run /help in Claude Code.

Delegation

• After creating/modifying commands: Run uv run .claude/skills/meta-command-creator/scripts/validate-command.py <command.md>
• Pattern discovery: For existing command patterns, use Explore agent

References

• Official Docs: code.claude.com/docs/en/slash-commands.md
• Existing commands: .claude/commands/openspec/
• Skills comparison: code.claude.com/docs/en/skills.md