omk-control-loop-debugger

Debug multi-agent workflow oscillation by mapping the runtime to a control loop, comparing fresh versus stale sensor evidence to acceptance criteria to identify whether the controller decision or actuator path causes the instability.

Agent runtime misroutes tools in DAG execution when the controller decision mismatches the required skill, worker, or provider, requiring an evidence-backed correction to verify the proper actuator path and restore loop stability.

Fix retry and fallback loops in agent workflows by separating hidden disturbances from root causes using sensor evidence, then defining a correction that verifies actuator authority and eliminates the repetitive control error.

Diagnose evidence-gated failures across worker lanes by comparing current sensor evidence against acceptance criteria to detect drift, identifying the responsible controller decision, and applying a correction that demonstrates verified loop stability.

Stop a multi-agent runtime from losing context by treating context as sensor evidence within a control loop, identifying where stale evidence causes bad controller decisions, and enforcing an evidence-backed correction plan.

Use a control loop framing for agent debugging when workflows exhibit oscillation, saturation, sensor drift, or controller mismatch, allowing you to separate root causes from symptoms and produce an evidence-driven correction plan.