{"agent":"claude","input":"Review failed n8n executions from 2026-11-12T06:00:00Z to 2026-11-13T06:00:00Z. Call get_instance_context once. Then search_workflow_executions with status error and crashed in that window, following nextCursor. For each execution call get_workflow_execution with includeData true and truncateData true, and find the node that stopped and its error message. Group failures by workflow and cause. For each group: say whether a later execution of that workflow succeeded, and whether get_workflow_history shows a version saved in the 48 hours before the first failure (if so, summarise get_workflow_versions_diff between it and the version before). List the order numbers, SKUs or other ids in each failed execution's input. Write failures.md, most orders affected first. Do not call execute_workflow, test_workflow, publish_workflow or unpublish_workflow, and change nothing in n8n.","connections":["n8n"]}