Skip to content
TakoAPI
Back to skills

moa

by jscianna

AI & LLMs
367Downloads reported by clawskills.sh
0Views
0 likesSign in to like this skill
—Not yet rated
Your rating

Brief

Mixture of Agents: Make 3 frontier models argue, then synthesize their best insights into one superior answer.

Description

What This Skill Does

Mixture of Agents (MoA) sends a prompt to 3 frontier models in parallel, then passes all three responses to a fourth model that synthesizes them into a single answer. Costs ~$0.03 per query on the paid tier using OpenRouter.

Running three models in parallel and synthesizing their outputs catches blind spots that any single model, including GPT-4 or Claude Opus, would miss, at a lower total cost than a single Opus call.

When to Use It

  • Evaluating startup competitive moats before an investment decision
  • Comparing technical approaches like RLHF vs DPO for alignment
  • Cross-referencing factual claims across models with different training data
  • Generating a market landscape overview with diverse model perspectives
  • Getting a second and third opinion on a high-stakes architectural decision

Example Workflow

INPUT

User asks: Analyze the competitive landscape for AI code generation — who has defensible moats?

AGENT
  1. 1

    Send the prompt simultaneously to Kimi K2.5, GLM-5, and MiniMax M2.5

  2. 2

    Wait for all three proposer responses (parallel, ~45s)

  3. 3

    Pass all three responses to the aggregator model (Kimi K2.5)

  4. 4

    Aggregator resolves conflicts and synthesizes the strongest insights

OUTPUT

A structured analysis combining unique perspectives from all three models, with conflicting claims resolved and the most defensible points surfaced

Requirements

  • OPENROUTER_API_KEY environment variable (obtain at https://openrouter.ai/keys, paid tier required for default models)

Installation

clawhub install jscianna/moa

Run the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.

Links