I told it "Here is a shell prompt, try running a command..." then copied its command into the shell, copied the output back into ChatGPT, with the prompt "Here's the output, try another" and so on.
I did the same yesterday. don't forget to add the following to the prompt: "for dangerous commands, such as `rm` emit a "danger!" warning.
if such a warning occurs, I insert a confirmation dialogue.
Next: parse and compile the code and language in
```language
blocks
```
I haven't tried this with recent versions, but this was great fun early on: