Do I need to approve every step of the plan?

Not necessarily. The skill uses adaptive gating. If the AI's confidence score is above 95%, it can proceed automatically; for lower confidence levels, it will pause and ask for your explicit feedback or selection.

What is the consensus matrix algorithm?

It is a mathematical scoring system that clusters similar ideas from different models. It calculates an agreement score (how many models agree) and a confidence score (weighted average) to identify the most reliable solution.

What happens if one of the AI models is down?

The skill features built-in fallback chains. If a primary explorer model like Gemini fails or times out, the system automatically attempts the request with the next model in the chain (e.g., DeepSeek or Sonnet) to ensure the workflow continues.

How does the multi-model exploration work?

The skill sends your requirements to several LLMs (such as Claude Opus, Gemini, and Grok) simultaneously. This parallel execution generates diverse perspectives and technical approaches much faster than sequential prompting.

Resilient Multi-Model Planning

Name: Resilient Multi-Model Planning
Author: MadAppGang

byMadAppGang

•

230

•

生産性とワークフロー

Orchestrates multiple AI models to generate, score, and validate architectural designs through a resilient consensus-based workflow.

This skill transforms high-level ideas into production-ready designs by leveraging a diverse ensemble of AI models—including Claude, Gemini, and Grok—to explore solutions in parallel. It utilizes a sophisticated consensus matrix algorithm to identify the strongest architectural patterns, calculates confidence scores based on model agreement, and employs adaptive gating to ensure human oversight at critical decision points. By implementing robust fallback chains and structured task management, it provides a highly reliable environment for complex technical planning and complex problem-solving.

主な機能

01Resilient fallback chains ensuring planning completion even if specific model APIs fail

02Structured phase-based workflow from initial problem analysis to final implementation validation

03Automated consensus matrix algorithm that clusters and scores similar technical approaches

04230 GitHub stars

05Adaptive confidence-based gating that determines when to proceed automatically or seek user approval

06Parallel multi-model exploration using Claude, Gemini, and Grok for diverse ideation

ユースケース

01Brainstorming resilient API designs and rate-limiting strategies

02Designing complex system architectures and evaluating technical trade-offs

03Creating detailed, step-by-step implementation plans for new feature integrations

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add madappgang/claude-code planning

For use in Claude.ai and ChatGPT

Download Skill