7 out of 7 tests pass. That number is boring and exactly correct. Adding a CLI subcommand sounds trivial until you count the actual failure modes: the subparser is not wired up, the func is not registered, the dry run flag is silently swallowed, the argument parser crashes with a useless error. Any of those can bite you. The only way to know they do not is to test at three distinct levels, which is what this commit actually does. Here is what wiring warmup eval looks like in practice. ...