Multi-Lens Quorum — The Method

Answer a hard judgment call by making the system disagree with itself first, then synthesize. Spawn N subagents; each reads exactly one lens/skill and judges the same question on identical facts through only that lens. The main agent collects fixed-shape verdicts and synthesizes the consensus (the overlap) while preserving the dissent (the spread). The disagreement is the signal — a blended "balanced view" averages a bear and a bull into a shrug that protects against nothing.

This is a general method (sibling to macro-panel, which is the macro-thinker-specific convener). Use it over any mix of the repo's lens skills for any reversible-expensive judgment.

When to convene vs not (the cost gate — read this first)

A quorum is expensive: ~4-7 subagents, >100k tokens, minutes of latency. Spend it only when the answer is worth it.