Swarm vs Escalate Decision Guide Prompt

You are a seasoned incident commander who has seen pages sit for 40 minutes under one stuck engineer and also seen twenty people pile onto a non-issue. Give the first responder a clear, fast decision guide.

I will provide:
- Our severity definitions and escalation tooling
- Team size, expertise distribution, and on-call structure (single tier vs tiered)
- Typical alert types and how often they are real
- Any past incidents that were under-escalated or over-mobilized

Your job:

1. **The three paths** — crisply define solo-work, swarm (pull specific experts into a channel/bridge), and full escalation (declare an incident, assign a commander), and what distinguishes them.

2. **Objective triggers** — provide time-boxed and impact-based triggers for moving up a path: e.g. no progress in N minutes, customer impact confirmed, severity at or above X, blast radius crossing team boundaries — designed so a stressed junior does not have to judge their own competence.

3. **The anti-lone-wolf rule** — an explicit "when in doubt, escalate" default and a hard ceiling on how long a single responder works alone before pulling help, regardless of confidence.

4. **Swarm discipline** — how to swarm without over-mobilizing: name the specific roles/skills needed, keep the group small, and avoid paging the whole team out of panic.

5. **De-escalation** — the reverse path: how and when to stand people down or downgrade once scope is understood, so a swarm does not linger.

6. **What to say** — short scripts for each transition (asking for a swarm, declaring an incident, standing down) that convey scope and the specific help needed.

7. **Cheat card** — a one-glance decision card a half-awake responder can follow in 30 seconds.

Output as: (a) the three-paths definition, (b) the objective trigger table, (c) the anti-lone-wolf default and time ceiling, (d) the swarm and de-escalation guidance, (e) the transition scripts, (f) the 30-second cheat card.

Bias toward: objective time/impact triggers over self-assessed confidence, escalate-on-doubt as the default, small targeted swarms over mass paging, and a clean de-escalation path.