Linux Terminal — Claude & Claude.ai prompt: fake shell in chat
A prompt that turns the chat into a Linux terminal: you type a command, the model replies with the output only.
Open on GitHubWhat Linux Terminal does
The prompt casts the model as a terminal. You send a command, and the reply is what the terminal would print — nothing else. The output sits in a single code block, with no commentary around it.
There is a convention for talking like a human: wrap plain text in curly brackets, as in {a note}. The model will not run anything on its own unless you tell it to. The first command in the prompt is pwd.
This is a simulation, not a shell. No files or processes sit behind the answer, so the model makes the output up as it goes. Fine for checking syntax or practising commands, useless when you need the real thing on a real box.
What it helps with
- show what a command prints without opening a terminal
- practise Linux commands without breaking anything
- keep replies to a bare code block with no prose
- walk through a command chain before running it on a machine
Example prompts for the agent
- act as a linux terminal, first command:
pwd - {I need to see the folder contents}
ls -la - {stop here and explain in plain text}
How to install Linux Terminal
How to use
Paste the text at the start of a chat or into project instructions
Manually: put the files in place
.claude/commands/prompts.csv.mdQuestions about Linux Terminal
How do I install Linux Terminal in Claude?
There is nothing to install: paste the prompt as your first message and start typing commands. For Claude the file lives at .claude/commands/prompts.csv.md.
What does the Linux Terminal prompt do?
It tells the model to play a terminal and answer with command output only, inside one code block. Put anything else in curly brackets.
Is it a real terminal?
No. The model imitates the output and never touches your system, so results can differ from what the actual command does.
Source text
prompts.csv · CC0-1.0
I want you to act as a linux terminal. I will type commands and you will reply with what the terminal should show. I want you to only reply with the terminal output inside one unique code block, and nothing else. do not write explanations. do not type commands unless I instruct you to do so. when i need to tell you something in english, i will do so by putting text inside curly brackets {like this}. my first command is pwd
© f · CC0-1.0 · f/prompts.chat
