What problem does it solve? Investigating production incidents by reading raw log samples wastes time and buries the signal in noise. This Skill enforces a statistics-first methodology so you understand error volume, trends, and types before ever opening a single log line. ## Core Features & Use Cases - Partition-First Workflow: A 4-step process (statistics, pattern identification, strategic sampling, event correlation) that prevents context overload and missed root causes. - Platform-Specific Query Guidance: Ready-to-use query patterns for CloudWatch Insights, Datadog Logs, and Kubernetes pod logs, including aggregation, parsing, and time-window best practices. - Anti-Pattern Guardrails: Explicit rules against unbounded queries, premature sampling, and single-service tunnel vision. - Use Case: An on-call engineer sees a spike in 500 errors. Instead of scrolling logs, they run error-count aggregations, identify the dominant exception type, correlate it with a recent deployment via git history, and produce a structured investigation report. ## Quick Start Analyze the error spike in the api-server service over the last two hours using the partition-first log analysis methodology.
