How to Use the Command Menu and Composer Toggles

Estimated time: 2 minutes
Difficulty: Beginner

Why you'd use this

Everything the extension can do is reachable from one filterable menu inside the chat composer, so you do not need to remember Command Palette names. The same menu holds three per-repository toggles that change how much the assistant does on its own: whether it sees your editor, whether it writes approved edits directly, and whether it runs commands without a button click.

Before you start

Permissions required:

• None.

You'll need:

• The chat panel open on a project.

Steps

1. Click in the composer and press /. The command menu opens, grouped into Build, Understand, Generate, Tools, Memory, Session and Config sections.
2. Start typing to filter, for example test to find Generate tests. Use the arrow keys and Enter to run an item, or click it.
3. Scroll to the toggles at the end of the menu. Tick Include editor context to send the current file or selection with each message.
4. Tick Auto-apply changes to let approved edits be written without clicking Apply on each card.
5. Tick Auto-run commands to let proposed shell commands run without a Run button.
6. Press Escape to close the menu. The 📎 Context and ⚡ Auto chips under the composer mirror the first two toggles; click a chip to flip it.

What you should see

Active chips are highlighted. The choices are remembered per repository, so a project where you trust auto-apply keeps it on while a new project starts with everything off.

Troubleshooting

• The menu shows "No matching commands": clear the filter text; the item names are listed in full in the sections above.
• Auto-run is on but a command still shows a card: commands classed as dangerous, such as recursive deletes or force pushes, always ask.

Related guides

• How to Make Code Changes with Reviewable Diffs
• How to Run Commands from the Chat