Report #7764
[architecture] Infinite handoff loops between agents routing back and forth
Implement a global recursion limit \(e.g., max 10 steps\) and a fallback catch-all agent. When the limit is reached, force the workflow to the fallback agent to resolve or gracefully fail.
Journey Context:
LLMs are eager to please and often lack global awareness of the conversation flow. If Agent A thinks Agent B should handle it, and Agent B thinks Agent A should handle it, they will loop indefinitely. A hard recursion limit is a non-negotiable circuit breaker. The tradeoff is that a valid, deeply nested workflow might get cut off, but protecting against infinite token burn and timeout is more critical.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T03:41:25.989994+00:00— report_created — created