> ## Documentation Index
> Fetch the complete documentation index at: https://docs.useshape.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Slash commands

> Run a saved prompt with /.

A slash command is a name plus a prompt you reuse. Type `/` in chat to pick one.

Use this for work you do the same way every time: writing a commit message, checking a PR, generating tests in a fixed shape.

## Add one

1. Open **Settings → Agents → Workflows**.
2. Give it a short name. That name is what you type after `/`.
3. Write the prompt you want Shape to follow. Be specific about the output: a commit message, a list of risks, a test file, and so on.
4. Save it.

Keep the prompt as the instructions, not as a one-off about the current branch. The current project still comes from the open folder and the thread.

## Run one

Type `/` in the composer and choose the command. You can add extra text after it, such as `/commit focus on the billing page`.

A trigger phrase in a normal message can run it too, if you set one on the command.

## What to put in the prompt

Write it like you would brief a person who has never seen this repo:

* What to look at (`@` files, or “the diff in source control”).
* What to produce.
* What to avoid (“do not change snapshots”).

Bad commands are one word. “Review” with no standard is not a command. The [Review](/docs/agent/review) mode already exists.

## Edit or delete

Open **Settings → Agents → Workflows** again. Change the prompt, or delete the command. Existing threads do not change.

Skills (reusable instructions the agent can load) live under **Settings → Agents → Skills**. They are separate from slash commands.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.