slash-commands

# Slash Commands ## Purpose Turn a workflow you repeat into a command you invoke. A slash command is an explicit, user-triggered prompt — as distinct from a skill, which the model decides to load on its own. ## When to Use - A prompt you find yourself retyping. - A multi-step workflow with a fixed shape: review, release, triage, summarize. - A team workflow that everyone should run the same way. - Anything you want to invoke deliberately, rather than have the model choose. ## Capabilities - Command authoring with arguments. - Composing tool calls and shell commands into a workflow. - Scoping: project-level, personal, or bundled in a plugin. - Distinguishing command from skill. ## Inputs - The workflow, and the variable parts of it. - What the user will type, and what they will pass as arguments. ## Outputs - A command file that runs the workflow reliably. - Clear argument handling with a sensible default. ## Workflow 1. **Decide: command or skill?** — A skill is loaded when the model judges it relevant. A command is invoked when the user decides. If you want control over *when* it runs, it is a command. 2. **Name it for the action** — `/review`, `/release`, `/triage`. A verb the user would think of. 3. **Handle the arguments** — What varies between invocations? A PR number, a file, a version. Provide a default where one is sensible. 4. **Write the workflow as instructions** — The command file is a prompt. It should say what to do, in order, with what constraints. ...