A Technical Paper on State Machine-Based Workflow Automation Abstract This paper presents a declarative YAML-based workflow system designed to guide AI agents through complex multi-step tasks with deterministic state transitions. The system replaces brittle "nudging" approaches with explicit state machine architecture, providing clear instructions, tool usage constraints, and value capture mechanisms at each workflow stage. We describe the system's architecture, YAML schema design, state machine engine implementation, and demonstrate its application through a bug investigation and fix workflow.