The guardrails · Incidents
Crisis Mode
When production breaks
When production breaks, the week pauses for the fix. Crisis Mode is a short, practiced path – restore first, understand second, learn that Thursday afternoon.
| Severity | What it means | What happens |
|---|---|---|
| Sev 1 | Customers can't use a core flow, or data is at risk | Crisis Mode, now |
| Sev 2 | A core flow is degraded for many customers | Crisis Mode, same day |
| Sev 3 | A minor issue with a workaround | Into the live queue of customer concerns, reviewed at The Collab |
- 1
Minutes · Engineering
Stop It
Flag off, or roll back – questions later. Restore customers first. One-click rollback exists for exactly this moment.
- 2
Engineering
Declare It
Name the severity, and open a Crisis Intent. A Crisis Intent skips Intent Review. Engineering leads it, CX owns what customers are told, and Product keeps the business informed.
- 3
The IDE · the Pipeline
Fix It
Fast, but never around the gates. The fix is built in the IDE and goes through the Pipeline like everything else. Gates are never skipped, even under pressure.
- 4
The week
Pause What Clashes
The week bends, it doesn't break. If Coding Day needs the same people, it pauses. A same-day Final Review can be held just for the fix.
- 5
Thursday
Adjust KPI Learnings
Don't let a crisis skew the numbers. KPI Learnings for the affected area are marked, and left out of that week's calls.
- 6
Thursday PM
Learn
The session writes the incident review. AI generates the incident review from the session. The team reviews it at The Collab, and the lesson goes into the guardrails and context files.