/loop

The skill executes an iterative refinement loop by running a task, evaluating the output against a defined quality condition, and repeating until the condition is satisfied or max iterations are reached.

Quality-aware termination stops the task processing loop when a defined quality criterion is met. If the condition is not satisfied, the loop continues using the prior output as a base until the maximum iteration limit is reached.

Progressive improvement is achieved using an end-to-end refinement loop that evaluates each output against a quality condition and incrementally addresses issues across successive iterations to achieve the desired result.

The loop runs a default of 5 iterations for code or content refinement if a specific maximum is not provided. It stops earlier if the defined quality condition is satisfied during the task processing.

Yes, you can observe iteration logs during the refinement process. The loop implements end-to-end refinement with observable iteration logs and a final output, allowing you to track progressive improvements across each cycle.

You should not use an iterative refinement loop when a task requires only a single execution or lacks a clear quality condition to evaluate against. Without a defined stopping criterion, the loop will run until the default maximum iterations are reached.