poly-consultlisted

# Poly-Consult — multi-node consultation with summa synthesis > A single LLM node carries its own training bias and context frame. > For decisions that propagate through the system (A14 cascade), a > single viewpoint reproduces that bias at every downstream node. > Consulting N independent nodes, then synthesizing the responses into > a summa, surfaces what each alone could not: convergence (shared > confidence), dissensus (real uncertainty), and emergence (insights > that arise only at intersection). ## The principle Two observers of the same surface from different angles see different structure — intersections reveal what lies perpendicular to any single angle. This is geometry, not ideology. The skill formalizes that geometry for decisions. The protocol has three operations: 1. **Dispatch**: send the same question + context to N nodes. Each responds without reading the others. Isolation is the constraint that prevents the first responder from anchoring the second. 2. **Synthesize**: extract three structures from the N responses — agreement, disagreement, emergence. Output is not a vote, it is a matrix that preserves dissensus rather than averaging it away. 3. **Escalate** (optional, recursive): send the summa back to the same nodes (or expanded set) and ask the second-order question: "seeing this summa, does your first answer change?" Iterate to stable state. f(f(x)) applied to multi-agent decisions. ## When this skill activates **Trigger signal