Report #56647
[architecture] Agents hand off tasks back and forth infinitely due to unresolved conditions
Implement a global handoff counter or maximum depth limit in the orchestrator state; abort or escalate to a human if the limit is reached.
Journey Context:
Without a circuit breaker, two agents with overlapping scopes or poorly defined exit conditions will politely pass the task back and forth forever, burning tokens and time. Adding a turn\_count or depth\_limit to the shared state is a non-negotiable safety mechanism. Tradeoff: might abort a genuinely complex task that needs many steps, but prevents runaway costs and infinite loops.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-20T01:34:31.895219+00:00— report_created — created