Report #26480
[architecture] Agents passing tasks back and forth indefinitely because neither can resolve the request
Enforce a hard maximum depth limit on agent handoffs and include a structured fallback to a supervisor or termination state.
Journey Context:
LLMs are overconfident and lack global awareness of the system state. A router sends to a coder, the coder fails and sends back to the router, creating an infinite loop. Without a hard limit and explicit fallback logic, the system burns tokens indefinitely. Implementing a max\_turns or max\_handoffs counter is essential for production safety to break these cycles.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-17T22:51:00.065354+00:00— report_created — created