Prefect Flow Orchestration AI Skill Guide

Overview & Engine Architecture

Prefect 2/3 wraps Python functions as @task / @flow with state tracking in Prefect API/Cloud. Deployments schedule flows onto work pools/workers. Agents keep tasks pure and idempotent, configure retries/caching deliberately, and pass artifact URIs instead of giant in-memory payloads between tasks.

@flow / @task code
      -> Prefect API (runs, states)
          -> work pool / worker
              -> infrastructure (process, Docker, K8s)

When to use this skill

• Python-first pipelines with lighter DAG boilerplate than Airflow
• Deploying the same flow to local, Docker, or Kubernetes workers
• Retries, caching, and run observability for ETL/ML batch jobs