agent-skills

Use when the user reports a GlassFlow pipeline failure, missing events, or unexpected behavior. Starts with component logs (via kubectl or whichever log backend the OTel Collector routes to) and only reaches for metrics when the logs are clean or ambiguous. Triggers on phrases like "pipeline is failing", "no events arriving in ClickHouse", "diagnose pipeline X", "why is pipeline Y stuck".

Use when the user asks to create a GlassFlow ETL pipeline. Walks through source configuration (Kafka or OTLP), optional dedup/filter/transform stages, and ClickHouse sink mapping, then submits the v3 config to the GlassFlow API. Triggers on phrases like "create a pipeline", "set up a Kafka to ClickHouse pipeline", "ingest logs from OTLP into ClickHouse".

Use when a GlassFlow pipeline is experiencing backpressure, sustained sink lag, or throughput pain. Reads GlassFlow's emitted Prometheus metrics to confirm the bottleneck (backpressure gauges, stream depth ratio, source vs sink rate, sink batch size distribution), proposes targeted changes to replicas, batch sizes, stream limits, or dedup storage, applies them via the appropriate API endpoint, and verifies improvement by re-reading the same metrics. Triggers on phrases like "pipeline has backpressure", "tune pipeline X", "throughput is too low", "size resources for pipeline Y".