Agent Loop
After a check fails, should the AI continue editing, stop to ask me, or end the task?
$ check mobile-nav ✗ The page still scrolls horizontally
Continue after an initial failure, pause at a limit or approval boundary, and stop after acceptance checks pass.
Continue after an initial failure: If the failure is new and attempts remain, the application passes the check result into another round.
Pause at a limit or approval boundary: Two identical failures or a proposed shared-style change pauses the loop with its evidence.
Stop after real checks pass: The task ends only when the build passes and the 390px page no longer overflows.
You can say this to an AI Agent
Fix the mobile navigation overflow. After each change, run the build and inspect the page at 390px. Allow at most two attempts for the same failure and pause before changing shared navigation styles. Finish only when both checks pass, and report why each round continued, paused, or completed.
Select page elements for precise AI edits
Click a button, text, or card, note the change, then copy its exact location to AI.
Open Selector Pro website↗