| name | command-development |
| description | This skill should be used when the user asks to "create a slash command", "add a command", "write a custom command", "define command arguments", "use command frontmatter", or needs guidance on slash command structure, dynamic arguments, or command development best practices for Claude Code plugins. |
Command Development
Guidance for creating Claude Code plugin slash commands.
What Commands Are
Slash commands are Markdown files containing instructions Claude executes when
invoked. Plugin commands live in commands/ and appear as
/plugin-name:command-name when the plugin is installed.
Critical Rule: Write for Claude
Command content is instructions to Claude, not messages to the user.
Correct:
Review the target path for security issues including SQL injection and XSS.
Report findings with file:line references and severity ratings.
Incorrect:
This command will review your code for security issues.
You will receive a report with vulnerability details.
File Format
commands/
├── validate-plugin.md # /meta:validate-plugin
└── audit-skill.md # /meta:audit-skill
Use kebab-case filenames. The filename (without .md) becomes the command name.
Frontmatter
Optional YAML frontmatter configures behavior:
---
description: Short summary shown in /help
argument-hint: <path>
allowed-tools: Read, Grep, Glob, Bash
---
Common fields:
| Field | Purpose |
|---|
description | Help text for /help |
argument-hint | Hint for expected arguments |
allowed-tools | Restrict tools for the command session |
arguments | Named arguments with descriptions |
Dynamic Context
Commands can inject runtime values:
$ARGUMENTS — raw argument string from invocation$1, $2 — positional arguments (when using arguments frontmatter)@path/to/file — include file contents in command context!`command` — run bash and inject output (use sparingly)
Authoring Workflow
- Define the outcome Claude should produce
- Write step-by-step instructions in imperative form
- Add frontmatter for description and argument hints
- Reference a skill when the command delegates to shared guidance
- Specify output format when reports or structured results are expected
- Update the plugin README Components section
Command + Skill Pattern
For complex workflows, keep the command thin and activate a skill:
When invoked:
1. Use $ARGUMENTS as the target path
2. Apply the plugin-structure skill validation workflow
3. Produce a severity-ranked findings report
Additional Resources
references/patterns.md — Reusable command patterns and examples
