Terminal Command Generation

Kodik can write a shell command for you without leaving the terminal. Describe what you want in plain language, and Kodik returns a command that fits the terminal you are actually typing in.

How it works

1. Focus the integrated terminal.
2. Press Ctrl+I (Cmd+I on macOS). An input appears at your cursor.
3. Describe the command, for example “find every TODO added this week”, and press Enter.
4. Kodik shows the generated command with a short explanation.

You can also run Kodik: Generate Terminal Command from the command palette.

Inserting vs running

Once a command appears, you choose what happens to it:

Key	Action
Enter	Insert the command on the prompt without running it, so you can read or edit it first
Ctrl+Enter	Run the command immediately
Escape	Dismiss the input and return to the terminal

Insert is the default on purpose: Kodik never executes a generated command unless you explicitly ask it to.

Terminal-aware results

Generation is given the terminal’s current working directory, shell, and operating system. A command generated in a PowerShell terminal uses PowerShell syntax; one generated in zsh uses POSIX syntax; and paths are relative to the directory you are actually in.

Destructive commands

When a command deletes data, rewrites history, or is otherwise hard to undo, Kodik marks it before you accept it. The warning combines the model’s own assessment with a local check for well-known irreversible operations such as rm -rf, git push --force, and git reset --hard.

Which model is used

Terminal generation uses the same model you have selected in Kodik chat, including local Ollama models and custom providers. It runs as a single standalone request and never adds anything to your chat history.

The one-time hint

The first time you open a terminal, Kodik shows a short hint that Ctrl+I generates commands. It disappears as soon as you start typing and does not come back. To turn it off entirely, set:

"kodik.terminal.commandHint.enabled": false