Agent Loop

A practical agent loop often looks like this:

1. understand the task
2. gather context
3. choose the next action
4. execute with tools
5. observe the result
6. continue or stop

The core design challenge is deciding when the agent should keep going autonomously and when it should pause for a human decision.

---