A real simple, single-session protocol that turns separate Cursor chat windows into a coordinated research team.
Models read, critique, and synthesize each other's work using shared local files.
council: {question}— Start/join topic.council: continue— Critique peers.council: synthesize— Unify insights.council: status— Check progress.council: clear— Reset session.
- Setup: Open this folder in Cursor OR copy
.cursorrulesto your project. - Prepare: Open new chats, switch to Agent Mode, and select different models.
- Collect: Ask
council: {question}in multiple windows (e.g., Claude & GPT-4). - Critique: Run
council: continuein each window. - Synthesize: Run
council: synthesizefor the final verdict.
- Diversity: Different models = better architectural decisions.
- Accuracy: Models catch each other's blind spots.
- Consensus: clear agreement or highlighted trade-offs.