1. Create a reports directory.
2. Launch independent judge agents in parallel to analyze the solution.
3. Judges produce independent assessment reports.
4. Judges debate and challenge each other's assessments.
5. Judges revise their assessments if convinced.
6. Check for consensus among the judges.
7. If consensus is achieved, generate a consensus report.
8. If no consensus after max rounds, report persistent disagreements.
SynapticSkills