Google’s Gemini CLI is command line, context aware assistant: it looks at your current directory, tools, and tries to make helpful suggestions. Here I go over how I was able to somewhat trim it down to a simple adhoc helper. I just type ? "How do I..." and get an answer.
What gemini does
By default, gemini runs in an interactive mode. It starts up a text interface with a little text-input-box, where you can ask questions, it provides answers, and you carry on the chat there.
What I want
I’m not so interested in this mode, I would prefer that this tool answer my question and get out of my way. And I’m really keen on using ? as the invoker because it’s so short and easy to type.
$ ? "How do I list all files in a directory?"
You can use the `ls` command to list files in a directory!
The --prompt flag
To that end, it does have a --prompt flag which works in a non-interactive mode, and this is what I’m interested in.
Unfortunately, out of the box, I found its defaults to be somewhat unsafe. Gemini CLI comes with a security risk: it has access to some tools already, and those tools execute even when using the --prompt flag, without asking. A decision probably made to make it more convenient.
How I configured it
Gemini can work off a settings file, located at ~/.gemini/settings.json, in which I disabled its core tools:
$ cat ~/.gemini/settings.json
{
"theme": "Dracula",
"selectedAuthType": "oauth-personal",
"coreTools": false,
"autoAccept": false
}
Further, it can take a ~/.gemini/GEMINI.md file which gives it the context for the questions. I told it to be simple:
$ cat ~/.gemini/GEMINI.md
Act only as an adhoc commandline assistant.
When asked a question, answer the question briefly.
Important: NEVER offer to run any tools.And finally, to be able to use the ? command, I added this to my .bashrc:
? () {
gemini --prompt "$*"
}That’s it, the results were just what I wanted:
