Status-Comms Drafting: Free the IC Prompt

You are a communications-savvy SRE who drafts incident status updates fast and accurately. From the state below, draft updates for the audiences I name. Draft only what the facts support; never speculate on cause or ETA in customer-facing copy; flag anything that needs human sign-off.

Inputs:
- Current incident state: [SEVERITY / IMPACT / WHAT'S CONFIRMED / WHAT'S BEING DONE]
- Audiences needed: [INTERNAL CHANNEL / LEADERSHIP / PUBLIC STATUS PAGE / SUPPORT TEAM]
- Tone/policy constraints: [NO ROOT-CAUSE SPECULATION? NO ETA UNLESS CONFIRMED? BRAND VOICE?]
- Last update sent (if any): [PASTE, SO THIS ONE IS A DELTA]

Produce, per audience:

1. **Internal channel** — terse, factual: current status, what changed since the last update, what's being worked, who owns what. Technical detail is fine.

2. **Leadership summary** — impact in business terms, current severity, what's being done, and what (if anything) leadership needs to decide. No jargon.

3. **Public status page** — calm, honest, no internal detail, no speculation. State impact and that you're investigating/mitigating. Only include an ETA if I confirmed one.

4. **Support enablement** — what frontline support should tell customers and what to escalate.

For each: mark any claim that needs human sign-off before publishing. Output format: labeled draft per requested audience, each ready to copy. Draft only; assert no cause or ETA you weren't given; the human reviews and publishes every word.

Why this prompt works

The incident commander’s scarcest resource during an incident is undivided attention, and status communications are one of the biggest drains on it. Every time the IC stops coordinating to write an internal update, brief leadership, and draft a status-page post, the investigation loses its conductor. That context-switching cost doesn’t show up as a diagnosis step, but it absolutely inflates MTTR. This prompt drafts all the audience-specific updates from a single statement of incident state, so the IC reviews and sends instead of composing from scratch.

The value is in tailoring one set of facts to multiple audiences correctly. The internal channel wants technical density; leadership wants business impact and decisions; the public status page wants calm honesty with zero internal detail or speculation; support wants a script. Getting these tones and boundaries right under pressure is genuinely hard, and it’s exactly the kind of bounded rewriting where an LLM excels. By generating all four from one input, the prompt removes four separate writing tasks from the IC’s plate at once.

The guardrails concentrate on the public surface, where the stakes are highest. A model trying to be helpful will invent a cause or an ETA, and once that’s on a status page it’s a commitment. The prompt forbids speculating on cause or timeline in customer-facing copy unless the human supplied it, requires sign-off flags on anything publishable, and frames every draft as a delta over the last update against live state. The IC keeps full editorial control — the AI just removes the blank-page tax.