Agents / Agent Retry Storm Playbook

Agent Retry Storm Playbook

A step-by-step playbook to contain retry storms, protect dependencies, and restore stable agent behavior.

What matters in practice

Throttle first, then recover with explicit next-action ownership.
Keep one continuity contract for all runtimes and teams.
Track outcomes, not just incidents, to prove operational value.

Implementation checklist

Define stable session and handoff rules.
Instrument score, risk, and closure-rate signals.
Review results weekly and remove low-value loops.

Related