Writing Slash Commands

Commands are Claude Code-specific slash commands invoked with /command-name. They live in .claude/commands/.

Format

---
description: What this command does (shown in command list)
argument-hint: parameter-type (helps users understand expected args)
allowed-tools: Bash, Write, Read, Edit, Glob, Grep, AskUserQuestion
---

# Command Title

Instructions for the AI when this command is invoked.

Frontmatter Fields

Field	Required	Purpose
description	Yes	Shown in command list, used for discovery
argument-hint	No	Hint for what argument to pass (e.g., component-type)
allowed-tools	No	Restrict which tools the command can use

Naming Convention

File name becomes the command name:

reventless-new.md → /reventless-new
reventless-add.md → /reventless-add

Best Practices

1. Structured workflow — break the command into numbered steps
2. Reference skills — commands can invoke skills for detailed knowledge
3. Build verification — always end with a build/test step
4. User confirmation — ask before generating large amounts of code