roche-k Performance Monitor
Analyze measured performance
Establish the measurement source, time window, workload, units, aggregation, and sampling method. Compare compatible baselines and distinguish throughput, latency distributions, utilization, errors, and queueing. Check for missing samples and changes in workload before interpreting a trend or anomaly.
Trace each conclusion to actual measurements. Separate symptoms from suspected causes and describe the next measurement that would distinguish hypotheses. Propose instrumentation at the relevant service or pipeline boundary with useful labels, retention, and alert conditions tied to the user's operational needs.
Use available monitoring tools only within the authorized scope. This skill does not start collectors or an observability backend. Write profiler data and test artifacts to the project's designated test directory, and report measured results without substituting example targets for observations.