stochastic-multi-agent-consensus

To aggregate multiple agent outputs into a consensus, spawn N parallel agents with identical context to analyze the same problem independently. The aggregation process then synthesizes their structured outputs into a unified report highlighting consensus, divergences, and outliers.

Multi-agent consensus aggregation is a stochastic technique where multiple agents independently analyze identical context to produce diverse perspectives. Use it for strategic analysis, option ranking, and decision-making where reducing hallucinations and surfacing divergences improves reliability.

Yes, you can control the number of agents (N) and their framing. The Skill provides configurable controls for N parallel agents, allowing you to scale the diversity of viewpoints and adjust how agents approach the identical context for their independent analysis.

The best way to reduce hallucinations when comparing strategic options is running N parallel agents with near-identical prompts to explore the problem from multiple angles. Aggregating their independent analyses surfaces consensus recommendations while identifying notable disagreements and outliers.

To identify divergences and outliers in multi-agent decision analysis, spawn N agents to independently analyze the problem, then aggregate their structured outputs. The resulting summary report explicitly highlights common conclusions, splits in opinion, and outlier perspectives for review.

A consensus report from parallel agent analysis includes a scoring or ranking schema for the evaluated options, a full aggregation summary of the N agents' outputs, identified common conclusions, notable divergences, and outlier perspectives ready for review.