command-development

Slash commands for Claude Code are Markdown files with YAML frontmatter that define metadata, arguments, and instructions. Create a file under .claude/commands/ with frontmatter fields like description, allowed-tools, and model, then write your command logic below. This lets you automate workflows and standardize repetitive tasks across projects.

Slash commands support frontmatter fields including description, allowed-tools, model, argument-hint, and disable-model-invocation. These fields control help text, restrict tool access, set model behavior, document arguments, and toggle Claude's invocation. Proper frontmatter ensures commands behave predictably and integrate cleanly with Claude Code.

Yes, slash commands support dynamic arguments, file references, and inline Bash execution. Arguments are documented via argument-hint, file references let commands access project files, and embedded Bash enables scripting. This flexibility lets you build commands that adapt to different contexts and automate complex workflows.

Follow single-responsibility design: each command handles one task with a clear, verb-forward name like /git-commit or /review-pr. Store commands under .claude/commands/, use hyphens in filenames, and document purpose in the description field. Clear naming and organization make commands discoverable and maintainable across teams.

Use the allowed-tools frontmatter field to specify which tools your command can invoke. This limits scope, prevents unintended side effects, and hardens security by restricting access to sensitive operations. Define tools as a list to control what each command exposes to Claude Code.

No custom code is required. Slash commands are Markdown files with YAML frontmatter and plain-text instructions. You define the command's interface and behavior declaratively, and Claude Code interprets the metadata and executes the instructions. This design prioritizes simplicity and reusability.