What problem does it solve? When an alert fires on one microservice, the actual root cause often lives several hops downstream. Manually correlating traces, spans, and service dependencies in Jaeger is slow and error-prone, especially during cascading failures where the alerting service is only a symptom. ## Core Features & Use Cases - Trace Querying: Four tools (query_traces, get_trace_detail, get_service_dependencies, find_error_traces) to list traces, inspect full span trees, map service call graphs, and group error traces by origin service. - Guided Investigation Decision Trees: Step-by-step workflows for service degradation, intermittent failures, and unknown dependency failures, with span-tree interpretation rules (slowest span, error propagation, fan-out patterns). - Confidence Scoring: Structured evidence scoring that boosts root-cause confidence when traces show clear bottlenecks, repeated error origins, and metric correlation. - Use Case: An alert fires for high error rate on frontend. The skill maps dependencies, finds slow traces, drills into the span tree, and reveals redis-cart as the 4.2s bottleneck — not the frontend itself. ## Quick Start Ask the agent to investigate why the frontend service has high latency by analyzing recent Jaeger traces and identifying the root cause service.
