How the Production Engineering agent investigates and resolves real production issues — each one drawn from the actual ticket: the trigger, the evidence, the fix, the verification.
/prode-triageThe command that drives every case below: from a raw ticket to evidence, a surgical fix, a reviewed PR, and a clean Linear handoff.
A synthetic-monitor alarm caught an unguarded null-deref in the commitments endpoint before a real investor hit it.
A spurious 400 told investors their successful payment had failed, and spammed Sentry. Made idempotent.
A Live-Private deal showing to the "wrong" people. The agent proved it was working as intended and documented the rule.
A repeating RDS CPU alarm traced to five missing indexes on one admin search query.
A duplicate company entity was blocking a payment. Removed safely, without touching the correct entity's data.
Currency rates, pending payments, and the deals list made fast and resilient — before they trip an alarm.