Agentic Workflow

A closed-loop working mode in which an AI system autonomously breaks down tasks, formulates plans, executes actions, observes results, and iteratively corrects itself. In programming scenarios, this manifests as reading runtime errors, locating relevant code, generating fix patches, re-running tests, and deciding based on the results whether to continue modifying, without step-by-step human instructions.