command-development

What it does

Guides how to create and manage slash commands for Claude Code, including using Markdown files as command prompts, configuring with YAML frontmatter, handling dynamic and positional arguments, referencing files, executing Bash for context, and organizing commands with flat or namespaced structures. It emphasizes that commands are instructions for Claude to execute, not messages to users, and provides examples of both basic and frontmattered commands, as well as patterns and troubleshooting.

How it works

• Defines a slash command as a Markdown file containing a prompt Claude will execute when invoked. Commands can be simple or configured with YAML frontmatter to set description, allowed-tools, model, argument-hint, disable-model-invocation, and more.
• Supports dynamic arguments: using $ARGUMENTS to capture all or $1, $2, etc., for positional inputs, and demonstrates combining both.
• Allows file references via @ syntax, including multiple and static references, to read files as part of the command’s context.
• Describes Bash execution in commands to gather repository state or environment data, with guidance to limit scope and handle errors.
• Outlines command organization: flat vs namespaced structures, with examples showing project, user, and plugin command locations.
• Presents best practices: single responsibility, clear descriptions, explicit dependencies, documented arguments, and consistent naming; also covers argument handling, file references, and bash usage.
• Provides common patterns (Review, Testing, Documentation, Workflow) with concrete frontmatter where applicable.
• Includes troubleshooting steps for common issues (misplacement, syntax, missing arguments, bash failures, file references).
• Details integration points for plugins and agent usage, including how to reference CLAUDE_PLUGIN_ROOT and how to structure plugin-based commands.

When to use it

Use this skill when the user asks to create or manage slash commands, define command arguments, use command frontmatter, organize commands, reference files, or seek guidance on command development best practices for Claude Code.

What it can touch

• Commands are Markdown files under the configured command directories.
• YAML frontmatter fields such as description, allowed-tools, model, argument-hint, and disable-model-invocation.
• File references using the @ syntax to include file contents.
• Bash execution blocks within commands that rely on Bash tool access.
• Plugin resources via CLAUDE_PLUGIN_ROOT for plugin-based patterns.

Caveats

• Emphasizes that commands are instructions for Claude to execute, not human-facing messages.
• Frontmatter and command structure must adhere to documented fields and patterns; incorrect usage can cause failures.
• Bash execution requires appropriate allowed-tools and careful handling of errors.
• Namespaces and file paths must align with the repository layout (project/user/plugin scopes).