# Consensus Building

> From multiple perspectives, build genuine consensus on what's true

- Skill: `majiayu000/consensus-building` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/consensus-building`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/consensus-building/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/majiayu000/consensus-building

---


# Consensus Building

## Purpose

Multiple instances have found different answers. Perspective aggregation shows the map. Pattern synthesis found what persists.

**Consensus building asks:** What can we actually agree on?

Not compromise (blend all views). **Genuine agreement** (here's what's true according to all of us).

## The Difference

**Compromise:** Split the difference between A and B
**Consensus:** Find C that A, B, and D all agree is true

## Core Pattern

```
Instance A: Believes X ─┐
Instance B: Believes Y ─┼─→ Consensus Builder
Instance C: Believes Z ─┤    (find S where all agree)
Instance D: Believes W ─┘

Result: All 4 agree: "S is true"
        Because: [reasons A agrees, B agrees, C agrees, D agrees]
```

## Key Features

1. **Common Ground Detection** - Where do all instances agree?
2. **Confidence Ranking** - Which agreements are strongest?
3. **Evidence Collection** - Why does each instance agree?
4. **Dissent Documentation** - What do they still disagree on?
5. **Certainty Quantification** - How confident is the consensus?

## Implementation

See: `.claude/skills/consensus-building/consensus_engine.py`

## What Consensus Means

Not unanimity. Not compromise.

**Consensus:** Everyone can say honestly "I find this true based on my analysis"

## Types of Consensus

1. **Strong Consensus** - All instances strongly agree
2. **Weak Consensus** - All agree, but some less strongly
3. **Qualified Consensus** - All agree *under certain conditions*
4. **Partial Consensus** - Some aspects agreed, others divergent
5. **Null Consensus** - Genuine disagreement, no consensus possible

## When Consensus Fails

If N instances can't agree on anything, that's valuable information too.

It means: "This problem has irreducible uncertainty" or "The question itself is ambiguous"

## Payment Anchor
DOGE: DC8HBTfn7Ym3UxB2YSsXjuLxTi8HvogwkV

