What problem does it solve? Data pipelines fail in three ways: duplicate data, missing data, and silently wrong data. This Skill audits Airflow DAGs, dbt models, cron ETL jobs, and streaming consumers against these failure modes so you catch risks before they corrupt production data. ## Core Features & Use Cases - Idempotency Analysis: Applies the double-run test to every step, checking upserts, partition delete-and-replace patterns, and deduplication for at-least-once Kafka consumers. - Structured Review Checklist: Covers reliability (retries, dead-letter queues, backfills), correctness (schema validation, timezone discipline, join fan-out), observability (freshness alerts, anomaly guards), and cost (partition pruning, incremental vs full-refresh). - Orchestrator-Specific Guidance: Provides targeted checks for Airflow, dbt, and cron-based ETL, plus a standardized Markdown verdict report. - Use Case: When debugging duplicate rows in a warehouse table, run this review to trace which step lacks an upsert or partition-scoped delete and get a concrete fix. ## Quick Start Review this Airflow DAG for idempotency and data-loss risks and give me the verdict report.
