What problem does it solve? Designing and operating observability pipelines across metrics, logs, traces, and profiles involves dozens of decisions—vendor selection, collector topology, sampling strategy, cardinality control, and compliance—that are easy to get wrong and hard to debug. This Skill classifies your intent and routes you to the correct guidance, vendor category, or forensics playbook instead of generic advice. ## Core Features & Use Cases - Intent-based routing: Classifies requests into setup, migrate, investigate, alert, trace, tune, or route, then dispatches to the right resource guide across a 4-layer × 4-boundary × 7-signal coverage matrix. - Transport and pipeline tuning: Provides in-depth guidance on UDP/MTU thresholds, OTLP gRPC vs HTTP, Collector DaemonSet vs gateway topology, and tail-sampling recipes. - Incident forensics: Runs a 6-dimension localization flow (code, service, layer, host, region, infra) with cross-signal validation and vendor query syntax for Tempo, Honeycomb, Datadog, Jaeger, and more. - Use Case: When a 5xx spike hits your payment service in ap-northeast-2, invoke the investigate intent to acquire a trace_id from a metric exemplar, narrow the root cause to a Redis connection pool exhaustion in a specific service version, and trigger an automated canary rollback. ## Quick Start Ask the agent to set up an OpenTelemetry stack on Kubernetes or to investigate a latency spike in a specific service and region.
