Yeah, usually I'd steer my agents to never use the output directly from any command, and instead redirect it to a logfile, then force it to search/grep stuff directly from the log-file instead of just getting all the outputs at all times. Seems to work OK.