Report #65660
[architecture] Agents entering an infinite loop handing a task back and forth
Implement a hard maximum handoff counter and a fallback supervisor agent that catches exceeded limits.
Journey Context:
LLMs lack global awareness of the conversation graph. If Agent A delegates to B, and B's prompt says 'Ask A for X', they loop indefinitely. Adding a hard limit and a supervisor breaks the cycle. Tradeoff: might terminate a genuinely complex multi-step process early if the limit is too low, but infinite loops cost infinite tokens and completely block the system.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-20T16:41:25.949272+00:00— report_created — created