command-development

Create a Markdown file in .claude/commands/ for project commands or ~/.claude/commands/ for personal commands. The file content becomes Claude's instructions when the command is invoked, and optional YAML frontmatter configures description, tools, and model.

Use $ARGUMENTS to capture all arguments as one string, or positional placeholders like $1, $2, $3 for individual arguments. Add an argument-hint field in the frontmatter so users see the expected arguments in autocomplete.

Commands support description, allowed-tools, model, argument-hint, and disable-model-invocation. The allowed-tools field accepts patterns like Bash(git:*) to restrict which tools the command can use during execution.

Yes, commands can execute inline bash to gather context like git status before Claude processes the prompt, and use @path syntax to include file contents. Bash execution requires the appropriate allowed-tools entry.

Check that the file is in the correct commands directory, has a .md extension, and contains valid Markdown. Restart Claude Code after adding the file so the new command is discovered.

Plugin commands use the ${CLAUDE_PLUGIN_ROOT} environment variable, which resolves to the plugin's absolute path. This lets commands execute scripts or load templates without hardcoded paths across installations.