Zum Inhalt springen
TakoAPI
Zurück zu den Skills

moa

von jscianna

AI & LLMs
367Downloadzahlen laut clawskills.sh
0Aufrufe
0 LikesMelde dich an, um diesen Skill zu liken
—Noch nicht bewertet
Deine Bewertung

Kurzbeschreibung

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

Beschreibung

Was dieser Skill macht

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.

Wann du ihn einsetzt

  • 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

Beispiel-Workflow

EINGABE

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

AUSGABE

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

Voraussetzungen

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

Installation

clawhub install jscianna/moa

Führe den CLI-Befehl aus oder füge den Prompt in Claude Code, Cursor, Codex oder deinen Coding-Agenten ein.

Links