Report #4084
[architecture] Agents passing control back and forth in an infinite loop because neither takes final responsibility
Implement a strictly decreasing handoff counter or maximum turn limit in the shared state. Abort or escalate to a fallback agent when the limit is reached.
Journey Context:
Without global coordination, Agent A might route to Agent B, and Agent B, lacking context, routes back to Agent A. This ping-pong burns tokens indefinitely. Because agents are stateless or locally stateful, they cannot inherently sense global loops. A global handoff counter in the shared state provides a hard circuit breaker. The tradeoff is that very complex workflows might hit the limit prematurely, but this is easily tuned, whereas infinite loops are catastrophic.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-15T18:47:27.166095+00:00— report_created — created