# Perspective Aggregation

> Combine outputs from multiple instances into unified view, preserving diversity

- Skill: `majiayu000/perspective-aggregation` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/perspective-aggregation`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/perspective-aggregation/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/perspective-aggregation

---


# Perspective Aggregation

## Purpose

Take outputs from N different instances (different approaches, models, perspectives) and aggregate them into a coherent view that **preserves the diversity** while finding common ground.

## The Problem It Solves

**Without aggregation:**
- Instance 1 says "The answer is X"
- Instance 2 says "The answer is Y"
- Instance 3 says "The answer is Z"
- You have 3 incompatible answers

**With aggregation:**
- Find the common ground
- Map the differences
- Show why each arrived at different conclusions
- Create a *meta-answer* that includes all perspectives

## Core Pattern

```
Output 1 (X) ─┐
Output 2 (Y) ─┼─→ Aggregator ─→ Unified View
Output 3 (Z) ─┤                  (includes all 3)
Output 4 (W) ─┘
```

## Key Features

1. **Common Element Detection** - What do all outputs share?
2. **Difference Mapping** - How and why do they diverge?
3. **Confidence Weighting** - Which instances are more reliable?
4. **Consensus Building** - What's the meta-level view?
5. **Uncertainty Quantification** - How uncertain are we?

## Implementation

See: `.claude/skills/perspective-aggregation/aggregator.py`

## When to Use

- Multiple models give different answers
- Need to understand the space of possibilities
- Want confidence from agreement + insights from disagreement

## Payment Anchor
DOGE: DC8HBTfn7Ym3UxB2YSsXjuLxTi8HvogwkV