Short answer: No—you don’t need to add them to prefect.yaml if they’re already set on the ECS work pool. Prefect 3.x will use the work pool’s variables by default. Only add them to prefect.yaml if you want to override the pool’s values for a specific deployment. Details: - Valid variable names for ECS pools are

execution_role_arn

and

task_definition_arn

(also

task_role_arn

if you need a task role). - Precedence: 1) Work pool defaults (lowest) 2) Deployment

job_variables

in prefect.yaml (override pool) 3) Flow run

job_variables

at runtime (highest) - You cannot “reference” pool variables from within prefect.yaml. To use the pool’s values, just omit those keys from the deployment. Example (inherits from the ECS pool; nothing hardcoded):

deployments:
  - name: my-ecs-deployment
    entrypoint: flows/my_flow.py:my_flow
    work_pool:
      name: my-ecs-pool
      # No job_variables here — will use the pool's execution_role_arn and task_definition_arn

If you need to override for this deployment:

deployments:
  - name: my-ecs-deployment
    entrypoint: flows/my_flow.py:my_flow
    work_pool:
      name: my-ecs-pool
      job_variables:
        execution_role_arn: "arn:aws:iam::123456789012:role/ecsTaskExecutionRole"
        task_definition_arn: "arn:aws:ecs:us-east-1:123456789012:task-definition/my-task:1"

If you prefer not to hardcode, you can inject via environment variables when building/deploying:

deployments:
  - name: my-ecs-deployment
    entrypoint: flows/my_flow.py:my_flow
    work_pool:
      name: my-ecs-pool
      job_variables:
        execution_role_arn: "{{ $EXECUTION_ROLE_ARN }}"
        task_definition_arn: "{{ $TASK_DEFINITION_ARN }}"

Want me to sanity-check your pool’s current variables or your prefect.yaml? Feel free to paste them (with any sensitive bits redacted).