Hey do you have any idea of a prefect prometheus exporter which goal was for example to monitor unhealthy agents?

We provide an automation for monitoring (and alerting about) unhealthy agents. Regarding Prometheus - I haven’t seen any user contribution on this but you could certainly explore that option if you are interested. The process will get much easier and more observable in Prefect 2.0 - you’ll be able to inspect and preview the work queue your agent is polling from even directly from your CLI