I recommend grouping your heartbeat checks into 3 categories. Morning routines run once at the start of your day: pull latest code, check for dependency updates, review overnight CI failures. Throughout-day checks run every 30 minutes: verify the dev server responds, confirm database connections, watch memory usage.
What about things that can't wait 30 minutes? Urgent triggers should fire an immediate notification. If a deployment pipeline fails or a critical service goes down, mark those lines with [urgent] in your HEARTBEAT.md. OpenClaw checks them on every heartbeat tick with instant alerting.