FYI I created a general alerting automation that runs on a cron schedule (every 15 minutes) to watch for failures, if it sees one, its sends a notification via Telegram/Slack. It is packaged largely as a deterministic automation, which keeps routine checks cheap and avoids unnecessary LLM calls unless a full message is actually needed.

What failures does it alert on?

Automation failures. It just check for any failed automations in the last hour.

Ah gotcha If automations are misbehaving is it possible that this one may misbehave as well?

Screenshot 2026-06-16 at 1.53.27 PM.png

Ah cool, that seems like a good catch I was thinking of various failure modes. e.g if the automation server is down, then the automation that runs every 15 min to check for failed automations would not schedule Wondering if there's datadog monitors we can setup for reliable alerting

Yea, I am going to assume the automation server isn't going down. If you are an enterprise, you probably have separate reporting on keeping instances up.

Could be more subtle too, like automation schedules but convo fails to start The automation you have seems cool too, would be interested in following along on the things it catches

Yes, this is very blunt, it just says if the automation started and ended, there are lots of failure points that it doesn't cover.

Aren't they all like that? Deterministic until conditions are fulfilled

What I am getting at with the deterministic language is that a lot of my automations start as conversations (non-deterministic). Here is am doing work, checking if the automations have issues outside of a conversation, so this way I can avoid any LLM token costs.

Yes, we've done a lot of work encoding that behavior into our skills by default: https://github.com/OpenHands/extensions/blob/main/skills/openhands-automation/SKILL.md