{"agent":"claude","input":"Use the Datadog connection. Call search_datadog_monitors for every monitor's status, threshold and alert condition, and search_datadog_events for monitor-alert events over the last 30 days. Write monitor-audit.md ranking monitors by alert count, marking ones that recovered within 5 minutes most of the time, ones that flapped and ones with no notification target, and suggest a new threshold or window for the top 10, based on the condition already on the monitor. Read only. Do not call create_datadog_monitor or any write tool."}