antigravity-core

Multi-AI collaboration for decision support coordinates multiple AI perspectives to enhance problem solving. It transforms a single AI into a panel of expert viewpoints, applying structured thinking protocols to surface diverse analyses and synthesize robust decisions.

To run a multi-perspective review for trading-system design, invoke the skill to simulate viewpoints like Conservative Trader and Aggressive Developer. Follow the frame, analyze, synthesize, and decision steps to structure the cross-checks and record the final recommendation.

Yes, multi-agent thinking works for cross-document reviews by facilitating the evaluation of design decisions from diverse expert viewpoints. It updates documents and records the rationale for future audits and onboarding.

You can save multi-perspective session outputs for continuity by writing the synthesized findings to the `.agent/context/session-<date>.md` file. This records the session rationale and decision trees for future audits.

The limitation of simulated multi-perspective analysis is that it relies on a single AI generating diverse viewpoints rather than querying separate models. It enforces structured thinking protocols, but the depth of disagreement is bounded by one model's reasoning capacity.