You could have a command that recommends commands and then you select them on a drop-down list.
Alternatively if the dataset is verified you wouldn’t need to worry about it running dangerous commands, since it doesn’t know any. Or you could have a list of verified commands that run automatically and any command not on that list requires confirmation.
But this is missing the point that most of the time I know exactly what command I want to run so adding a LLM Is quite useless. The reason so much of linux is still relying on commands is because for a lot of people (myself included) commands are quick and efficient.
You could have a command that recommends commands and then you select them on a drop-down list.
Alternatively if the dataset is verified you wouldn’t need to worry about it running dangerous commands, since it doesn’t know any. Or you could have a list of verified commands that run automatically and any command not on that list requires confirmation.
But this is missing the point that most of the time I know exactly what command I want to run so adding a LLM Is quite useless. The reason so much of linux is still relying on commands is because for a lot of people (myself included) commands are quick and efficient.