The composer agent keeps making terminal commands which are blocking. Since I can’t really see what is happening in the command, I don’t know if it is blocking, or a long running command. It would be nice if pop out terminal didn’t result in the AI continuing. When you pop out now, the AI can’t see the result of the command, so I can never actually use it.
So say the agent runs a terraform apply without auto-approve. I should be able to pop it out, approve it, then wait for it to run and the agent will continue when it finishes and have access to the results of the command