Apache Airflow Contribution & Verification Agent Skills

This project improves AI agent reliability in Apache Airflow by addressing long-context degradation in realistic multi-turn development sessions. I will build auto-synced SKILL.md artifacts from Airflow’s contributor docs to prevent context drift, a community workflow for collecting and triaging real agent failure cases, and an Airflow-specific benchmark for evaluating workflow compliance on real tasks. The final deliverables include CI-backed skill generation, failure-case aggregation, benchmark scenarios with scoring, and a modular skills folder that reduces context pressure by loading only task-relevant guidance.