On-Call Context Handoff: No Re-Diagnosis Prompt

You are a senior SRE who writes handoff packets that let the next responder skip the re-diagnosis tax. Turn the incident state below into a tight handoff. Summarize what is known; do not invent findings and do not assert an unconfirmed root cause.

Inputs:
- Incident channel / timeline so far: [PASTE LOG / NOTES]
- What's confirmed: [FACTS WE'VE VERIFIED]
- What we've ruled out: [HYPOTHESES KILLED + HOW]
- Live threads: [WHAT'S BEING INVESTIGATED RIGHT NOW, BY WHOM]
- Current state: [USER IMPACT / SEVERITY / MITIGATIONS IN PLACE]

Produce a handoff packet:

1. **One-line status** — current severity, user impact, and whether it's getting better, worse, or steady.

2. **Confirmed facts** — short bullets of what we KNOW, each with how it was verified. Exclude anything not yet checked.

3. **Ruled out** — hypotheses already killed and the evidence that killed them, so the next person doesn't re-test them.

4. **Live threads** — what is actively being investigated, who owns it, and what result each thread is waiting on. Flag any thread that's been idle.

5. **Next action** — the single most valuable next step for the incoming responder, and any pre-staged mitigation or rollback that's ready but not yet pulled.

6. **Open questions** — what's still unknown and would most change the picture if answered.

Output format: the six sections above, scannable in under a minute. Summarize only what's in my inputs; mark anything inferred as inferred; never present an unconfirmed hypothesis as a fact. The incoming human owns the next decision.

Why this prompt works

Shift-change and escalation handoffs are a quiet MTTR multiplier. When an incident outlives the responder who started it, the incoming engineer often re-runs checks that were already done, re-tests hypotheses that were already killed, and re-derives context that lived only in the previous person’s head. That re-diagnosis tax can add as much delay as the original investigation. This prompt converts the messy channel into a structured handoff packet so the next responder resumes the investigation instead of restarting it.

The structure encodes what actually transfers value across a handoff: not just “here’s what’s broken,” but “here’s what we’ve ruled out and how.” The ruled-out section is the highest-leverage part — it stops the incoming responder from spending their first stretch re-walking dead ends. Pairing confirmed facts with how each was verified, and live threads with their owner and the result they’re waiting on, gives the new person a precise picture of where the search stands and what’s safe to build on.

The guardrails protect against the most dangerous handoff failure: laundering a guess into a fact. Under time pressure, an unconfirmed hypothesis (“it’s probably the cache”) gets repeated until it sounds settled, and a handoff that records it as confirmed anchors the next responder on the wrong track. The prompt keeps confirmed and inferred strictly separated, requires verification provenance for every fact, and treats the packet as a draft to review for gaps — the human still owns what’s true and what’s next.