Can I use MAGI EX for general software questions?

While it works for any query, it is most effective for open-ended creative tasks like feature ideation, naming, and strategic direction where 'three heads are better than one'.

How does MAGI EX utilize different AI models?

It spawns three sub-agents in parallel: the Scientist (native Claude Opus), the Mother (OpenAI Codex via MCP), and the Woman (Google Gemini via CLI) to provide independent viewpoints.

What is the purpose of the 'Debate' phase?

Debate mode allows each model to critique the others' proposals, helping the user understand the strengths and weaknesses of each option from different algorithmic perspectives.

Does this skill require external API configurations?

Yes, to use the full power of MAGI EX, you need the mcp-codex tool for OpenAI access and the Gemini CLI configured for Google model access.

How does MAGI EX handle the results of a session?

After consolidation and user selection, the skill can automatically trigger a 'writing-plans' handoff to turn the chosen brainstormed idea into a concrete development roadmap.

MAGI EX Brainstorming Panel

Name: MAGI EX Brainstorming Panel
Author: vinta

byvinta

•

113

•

Productivity & Workflow

Orchestrates multi-model brainstorming sessions using Claude, Codex, and Gemini to surface diverse project perspectives and creative features.

MAGI EX is a sophisticated orchestration skill for Claude Code designed to break through creative blocks and identify project blind spots. It assembles a panel of three distinct AI 'teammates'—the Scientist (powered by Claude Opus), the Mother (OpenAI Codex), and the Woman (Google Gemini)—to explore a single question or topic in parallel. By leveraging the unique reasoning patterns of different model families, MAGI EX provides a consolidated report of diverse proposals, trade-offs, and top picks. It includes an interactive debate mode for cross-model critiques and seamlessly hands off selected ideas to planning tools for immediate implementation.

Key Features

01Multi-model orchestration across Claude, OpenAI Codex, and Google Gemini backends

02Parallel agent execution with distinct 'Scientist', 'Mother', and 'Woman' personas

03Structured handoff to writing-plans skills for transition from ideas to action

04Automated proposal consolidation with deduplication and trade-off analysis

05Interactive Debate Mode for cross-model critique of proposed ideas

06113 GitHub stars

Use Cases

01Brainstorming 'what to build next' when a project roadmap feels stagnant

02Surfacing architectural blind spots by comparing different model family perspectives

03Generating creative feature ideas that a single AI model might overlook

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add vinta/hal-9000 magi-ex

For use in Claude.ai and ChatGPT