Opentelemetry

This is your reference when you need to instrument services with OpenTelemetry and actually get telemetry flowing to a backend. It covers the full stack: auto-instrumentation and manual spans for traces, metrics, structured logs with trace correlation, OTLP export, and Collector pipeline configuration including processors and tail sampling. The progressive disclosure keeps the entry point light (resource attributes, quick OTLP setup) but the full skill runs about 9000 tokens across concepts, semantic conventions, context propagation, and troubleshooting cardinality or sampling issues. Honest take: OpenTelemetry has a lot of moving parts, and this gives you the patterns to avoid the common pitfalls like breaking trace context or blowing up metric cardinality.