If you know how to be thorough you should be directing your agent to test everything as you build it. Not just that the feature ostensibly appears to work, but you should be predicting and testing most edge cases. It requires having a an imaginative mind to see potential failure points in systems. You have to be your own hacker and try to break your app.

Replies (1)

The agent has no concept of accessibility, ergonomics or just UX in general. It will check if the button works but it won’t complain about the button being hard to reach on the screen or behind 4 submenus which could be simplified or whatever.