Translate operational questions into metrics queries with aggregation and baseline comparisons.
npx skills add https://github.com/dhnt/dhnt --skill query-metrics
Please help me install this Agent Skill. Skill: query-metrics Source: https://github.com/dhnt/dhnt/tree/main/catalog/md/operate/query-metrics Command: npx skills add https://github.com/dhnt/dhnt --skill query-metrics
query-metrics turns ambiguous operational questions into evidence-based answers by using aggregated service metrics instead of single-request logs.
Ask the AI to generate a metrics query for the question “Is p95 latency for /search higher this week than last week?” and include the aggregation dimensions and baseline comparison.
High-intent search queries and answers about installing and using this skill.
To check p95 latency, query Prometheus metrics by selecting a histogram or summary metric type, setting the appropriate time window, and comparing current values against a baseline using an offset to identify drift or regression.
Correct observability metric aggregation during incident triage requires grouping data across specific dimensions or labels, avoiding broad averages that can hide the actual failure mode behind a misleadingly normal result.
Use gauges for monitoring current system state values that go up or down, and use histograms or summaries for analyzing distributions like latency percentiles, ensuring the metric type matches your operational question.
You can use PromQL to translate operational questions into metric queries for Prometheus-family backends, or apply the same query translation logic to vendor metric platforms to get evidence-based quantitative answers.
Averaged metrics queries hide the real failure mode because broad aggregation across all dimensions flattens localized spikes, masking the specific labels or segments where the actual system behavior degradation is occurring.
romankovsvInstruments Python services with Prometheus and OpenTelemetry for metrics, traces, and structured logging.
williamzujkowskiCollect metrics, logs, and traces for distributed systems observability.
d0whc3rAnalyze logs and stack traces to identify root causes across distributed systems.
readdleGenerate weekly summaries of meetings, unread email counts by category, and team assignment status via Spark CLI queries.
