What problem does it solve? Kafka application work fails at runtime when config keys, SQL syntax, or API surfaces are written from stale memory, and subtle pitfalls like non-co-partitioned joins or missing DLQs silently corrupt data pipelines. This Skill grounds every answer in official Confluent and Apache documentation and applies a checklist of known pitfalls to design, implementation, and review tasks. ## Core Features & Use Cases - Docs-grounded research workflow: Retrieves from docs.confluent.io, kafka.apache.org, Apache Flink docs, and context7 with deliberate version pinning, refusing unofficial sources. - Multi-surface implementation: Writes producer, consumer, Kafka Streams, Connect, ksqlDB, and Flink code/config with delivery semantics, schema compatibility, and error handling matched to the stated guarantee. - Design and diagramming: Covers topic naming, partitioning, key design, and retention choices, and produces data-flow diagrams via the archify skill or Mermaid fallback. - Use Case: Ask it to write an exactly-once producer in Java and it will verify current config keys against the docs, set acks/idempotence correctly, and leave a Testcontainers round-trip check proving it runs. ## Quick Start Ask the agent to design a Kafka consumer with a dead-letter queue for poison messages and verify the offset-commit strategy against the official docs.
