Back to resources
Engineering discipline
Severity-gated deployment for agents — the 24-hour rule
Why a green evaluation suite within the last 24 hours is the right pre-promotion gate, and how to operationalise it without slowing engineering velocity.
22 June 2026Miracle Alex
"Download PDF" opens your browser's print dialog — choose Save as PDF as the destination.
Why 24 hours
Shorter windows penalise small teams who legitimately cannot re-run a full harness on every push. Longer windows let yesterday's evaluation certify today's regressed code. 24 hours is the operationally honest middle.
Severity, not pass/fail
A blocker failure stops promotion. A regression failure raises an incident but does not block. A warning is trended. Engineering velocity survives; the safety floor does not move.