A familiar interaction with a coding agent looks like this:

"I implemented the feature, updated the tests, and everything passes."

That sounds reassuring.

But what exactly does everything passes mean?

Did the agent actually run the tests?