The worst pages are the ones you inherit blind. Shift ends, new shift starts, and the first page at 2 AM references a deploy nobody mentioned. Handover is where incidents are born.
The 6-line handover note (replace your template with this)
RISK: what is on fire or about to be (or "none")
DEPLOYS: what shipped since last handover + owner
WATCH: metrics/dashboards that look weird (not alerts yet)
CONFIG/FLAGS: anything changed that pages could attribute to
BLOCKERS: waiting on who, for what
CONTACT: who to wake, for what (exactly)
Six lines. Fills in 4 minutes. The rule that makes it work: no prose. Prose hides risk; lines expose it.
Why long handovers fail
Every multi-paragraph handover reads as "all clear." The reader skims, and the one dangerous sentence — the deploy that touched the payment path — is buried at paragraph 6. Lines with labels can't hide. If RISK is empty, the reader trusts the empty, because they scanned for it first.
Make it stick
- Handover happens in the shared channel, not a doc — the next shift must be able to search it at 3 AM.
- The outgoing engineer posts it; the incoming replies with exactly one word: "received." No reply = not handed over.
- Postmortems check the last 3 handovers. If an incident's cause was watchable, the WATCH line goes in the template retro.
If you want the handover templates, paging rules, and the full on-call kit, it's packaged here: Agent Ops 24/7 kit. More field notes at dev.to/hive80lab.
Top comments (0)