Apache Airflow — паттерны построения DAG

★ 8.5 · orchestration

airflow-dag-patterns is a Claude Code skill that guides building production-ready Apache Airflow DAGs with best practices for design, operators, sensors, testing, and deployment. It covers four core principles — idempotency, atomicity, incremental processing, and observability — and provides dependency patterns for linear, fan-out, fan-in, and complex task graphs. A ready-to-use ETL DAG example demonstrates default_args configuration, exponential retry backoff, cron scheduling, and catchup control, while targeted advice covers TaskFlow API usage and sensor reschedule mode to keep workers free. Extended pattern documentation is stored in `references/details.md`. The skill fits data engineers who are orchestrating batch pipelines, debugging failed DAG runs, or deploying Airflow in a production environment.