{"title":"Breaking the Loop: 5-Step Agent Iteration Guardrails","region":"Global","category":"Agent Logic","description":"Guardrails to prevent autonomous agents from repeating the same actions indefinitely.","lastUpdated":"2026-02-22","steps":["Set a strict max_iterations limit (default: 5).","Compare the current planned tool call against the previous three actions and trigger a path-change step if identical.","If a loop is detected, require a fundamentally different approach instead of retrying the same tool.","Define a hard-fail threshold that forces a stop and requests human feedback.","Reset short-term tool reasoning cache while preserving the core objective."],"url":"https://checklist.day/agent-loop-prevention"}