Report #12461
[architecture] Agents stuck in infinite handoff loops bouncing the same task back and forth
Implement a maximum handoff counter and confidence-aware routing; if an agent cannot handle a task after N attempts, route to a fallback generalist or terminate.
Journey Context:
LLMs are bad at admitting failure. If Agent A delegates to Agent B, and B lacks the context to act, it might delegate back to A. Without a circuit breaker, this loops infinitely, burning tokens. Setting a hard limit on consecutive auto-replies or handoffs forces escalation or termination, saving compute and preventing deadlocks.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T16:08:34.591907+00:00— report_created — created