Reaching the Goal

The most natural way for the loop to end is the most obvious one: the goal is actually reached. The file is organized, the ticket is resolved, the email is sent and confirmed. Once that's true, there's nothing left to decide or act on, so the loop stops.

Checking whether the goal is met happens during perceive: the agent looks at the current state and compares it against what the goal actually required. That comparison is what tells it whether to keep looping or stop — not a guess, and not just running out of steps to take.

This only works if the goal was actually checkable in the first place. "Organize this folder" is vague until it's turned into something an agent can verify — for example, every file sorted into a named subfolder. A goal that can't be checked against the state can't reliably tell the loop when to stop.

How does an agent determine that the goal has been reached?

A goal that can't be ____ against the current state can't reliably tell the loop when to stop.

Reaching the goal is checked during the perceive step, by comparing the current state against what the goal required.