First Ten Minutes
This chapter walks through what happens when you open CommandUI for the first time.
Launch
Section titled “Launch”The app opens with a single shell session. The header shows:
CommandUI v1.0.0 — ~/projectsThe terminal area occupies the left side. The plan panel on the right shows “No semantic plan yet.” The composer sits at the bottom with two mode buttons: Command and Ask.
Try a raw command
Section titled “Try a raw command”- Click the composer (or press
Ctrl+J) - Make sure Command mode is selected (
Ctrl+1) - Type
lsand press Enter
The command executes in your real shell. Output appears in the terminal. A history item is created with status success.
Try a semantic request
Section titled “Try a semantic request”- Switch to Ask mode (
Ctrl+2) - Type “show me changed files” and press Enter
The planner generates a command plan. The plan panel shows:
- Intent: your original words
- Command: the generated shell command (editable)
- Risk: low / medium / high
- Explanation: why this command was chosen
You can now:
- Run Plan — execute the command as shown
- Edit the command field, then Run Plan — execute your modified version
- Reject — dismiss the plan, nothing executes
- Save Workflow — store this command for future reuse
Explore history
Section titled “Explore history”Press Ctrl+H to open the history drawer. Every command you ran or rejected is listed with:
- Input text and source type (raw vs semantic)
- Status (success, failure, rejected, planned)
- Duration and timestamp
- Rerun, Copy, View Plan, and Save Workflow actions
Click any item to expand its details.
Open settings
Section titled “Open settings”Press Ctrl+, to open settings. Key options:
- Mode: Classic (minimal chrome) vs Guided (plan panel always visible)
- Default input: Command or Ask
- Confirm medium-risk: whether medium-risk commands need explicit approval
What to notice
Section titled “What to notice”After ten minutes you should have seen:
- A real shell that executes real commands
- An AI translation step that you fully control
- History that accurately reflects what happened
- The ability to edit any generated command before execution
- Risk indication on every generated plan