How checks are run, and what a pass means
The application runs in a separate test space, and each check passes or fails by what the application actually does.
Checks are run against the built application in a separate test space. A check passes only when the application does what was agreed.
Where they run
In a temporary test space with its own empty records. It is created for the run and removed afterwards. Nothing in it touches real records, and you do not need to set anything up.
What decides a pass
What the application actually does. Each check sends a request and compares the answer with what was agreed. The assistant’s opinion plays no part, and it cannot turn a failure into a pass.
When a check fails
LobLogix looks for the cause and proposes a correction to the application. The checks stay exactly as you locked them. A failure is never fixed by changing the check.
There are limits on how many times it may try, how long it may take and how much AI allowance it may use. When a limit is reached it stops and shows you where things stand.
When the specification is the problem
Sometimes a failure shows that what was agreed cannot work as written. That comes back to you as a proposed change to the specification, which you review like any other.
What a pass gives you
A version that passed every locked check is the only kind that can be installed for use. The version you install is the exact one that passed.