stuck-recovery

Stuck loop detection in automated workflows works by feeding task execution logs into an orchestrator, which then applies signals-based detection to repeated errors, edit loops, and agent tool loops. This produces structured advisory outputs to prevent productivity loss.

Recovering from repeated errors in agent retries requires advisory recovery strategies that prescribe specific escalation paths. The system detects the loop and generates advisory-only guidance to minimize downtime and manual intervention.

Advisory-only guidance for R010-style handoffs integrates directly with orchestrators to enable safe, stateful handoffs. It triggers context resets during loop detection to ensure workflow resilience without forcing automated actions.

This approach works across orchestration tasks, retries, and tool integrations by monitoring long-running tasks for repetitive failure patterns. It applies signals-based detection to identify when tools or agents become trapped in loops.

Automated workflows get trapped in repetitive failure loops due to repeated errors, edit loops, or agent tool loops. Signals-based detection identifies these patterns in execution logs and provides structured advisory outputs to guide recovery actions.