airflow

The airflow skill provides a comprehensive interface for data engineers and platform operators to interact with Apache Airflow 2.x and 3.x environments. By leveraging the af command-line interface, users can perform complex orchestration tasks directly from their AI agent environment. This skill is optimized for rapid troubleshooting and operational management of complex data pipelines, helping teams reduce the mean time to resolution for pipeline failures and deployment bottlenecks.

• Perform granular DAG operations including listing, exploring details, viewing source code, and toggling pause/unpause states.

• Execute and manage pipeline runs: trigger specific DAGs, clear task instances for retry, and monitor real-time execution statistics.

• Troubleshoot and debug: fetch task logs, diagnose failed runs, and identify parse or import errors within DAG files.

• Manage instance configuration: view global Airflow settings, list active connections, examine variables and pools, and audit installed providers or plugins.

• System monitoring: perform health checks, analyze API accessibility, and manage multiple Airflow instance contexts (e.g., prod, staging, local) seamlessly.

• Use this skill when keywords like 'Airflow', 'DAG', 'task log', 'import error', or 'broken DAG' appear in user requests.

• The skill acts as an abstraction layer over the Airflow REST API, providing agent-native tools for CLI-based management.

• Avoid using this for SQL-based metadata analytics; route such tasks to the analyzing-data skill instead.

• Ensure af is configured via environment variables (AIRFLOW_API_URL, etc.) or the ~/.af/config.yaml file for secure credential management.

• Always utilize the --dry-run flag when performing sensitive management operations like instance discovery to confirm changes before execution.