from Astronomer Agents302
Structured root-cause diagnosis for failed Airflow DAGs with actionable fixes, impact assessment, and prevention recommendations.
This skill equips an agent with a step-by-step Airflow DAG diagnosis workflow: identify failing runs, fetch task logs, categorize failure types (data, code, infra, dependency), assess impact, propose immediate fixes and long-term prevention measures, and provide ready-to-run commands. It is focused on pragmatic, operator-friendly remediation rather than abstract guidance.
Use this skill when you need a thorough root-cause analysis of failed Airflow DAG runs — for example: "diagnose and fix the pipeline", "why is this DAG failing", or when a task consistently errors after a deploy. It is appropriate for both single-run triage and recurring failure investigations.
af wrapper (uvx) with commands for logs, runs, health checks, and run management; explicit output structure (Root Cause, Impact, Immediate Fix, Prevention, Quick Commands).Best used by agents with CLI and observability access (Cursor/Copilot-style tooling or Claude Code with shell access) that can run af commands and inspect logs.
Airflow DAG debugging skill from Astronomer providing a structured 4-step diagnosis workflow (identify failure, get error details, check context, provide actionable output). Uses a CLI tool (astro-airflow-mcp via uvx) for interacting with Airflow. Well-written instructions with clear categorization of failure types and remediation steps. No scripts included — purely instructional SKILL.md.
Clean skill — no security concerns. Instructions reference a CLI tool (af/uvx) for Airflow interaction. No shell injection risks, no hardcoded credentials, no destructive commands. Well-structured markdown with clear steps. Lacks a references/ directory and scripts/ but doesn't need them for this type of instructional skill. Frontmatter is complete with name and description. Could benefit from more specific trigger patterns in description.
