Well, we can't consider that a bug as you've effectively made a command that targets the active window and follows the steps to delete a command right up through the confirmation
Just like if you had added all the actions to VA to open up Word and delete all the contents from your document - it's doing EXACTLY what you've told it to do.
Also, 'blocking potentially harmful profile actions' does not pertain to commands you create yourself - again, VA is doing EXACTLY what you are telling it to do.
I *can* look at ways to prevent this - if it's simple (that is, it's easy to code and it doesn't interfere with the normal operation of VA that users have become used to over the years), I can get it in right away. Otherwise, I cannot offer any kind of ETA as this is the very first time this has come up in like almost 10 years
A workaround for now would be to check the {ACTIVEWINDOWTITLE} token and see if it is, 'VoiceAttack' - bail out of the command if it is.