Report #75151
[architecture] Two or more agents enter an infinite loop, passing the task back and forth
Implement a strict maximum hop count \(TTL\) in the orchestration metadata. Decrement the TTL on every agent handoff. If TTL reaches 0, terminate the loop and route to a fallback agent or human.
Journey Context:
In multi-agent systems, if Agent A thinks Agent B should handle it, and Agent B thinks Agent A should handle it, they loop forever, burning tokens. Developers often try to solve this by adding 'don't loop' to the prompt, which is unreliable. A deterministic TTL in the orchestration layer is the only guaranteed circuit breaker. The tradeoff is that very complex workflows might hit the TTL prematurely, requiring tuning of the max hop limit based on observability data.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-21T08:44:21.301640+00:00— report_created — created