council-decision
Get multi-model AI consensus on complex questions using Claude, Codex, and Gemini
When & Why to Use This Skill
The Council Decision skill leverages a multi-model AI consensus mechanism involving Claude, Codex, and Gemini to provide high-quality recommendations for complex technical and architectural decisions. By utilizing a unique 3-stage voting and synthesis pipeline, it effectively reduces individual model bias and delivers diverse, peer-ranked insights to solve sophisticated software engineering challenges.
Use Cases
- Architectural Decision Making: Evaluate complex system structures by gathering consensus from multiple AI models to ensure robustness and scalability.
- Tech Stack Selection: Compare various technologies (e.g., SQL vs. NoSQL or WebSockets vs. SSE) with peer-ranked evaluations to choose the most suitable option for your project.
- Implementation Strategy: Determine the best coding patterns and error-handling strategies through a synthesized multi-model recommendation pipeline.
- Strategic Planning: Generate comprehensive multi-step plans for large-scale development tasks by aggregating diverse perspectives and strengths from different AI engines.
| name | council-decision |
|---|---|
| description | Get multi-model AI consensus on complex questions using Claude, Codex, and Gemini |
Council Decision Skill
Get multi-model AI consensus on complex decisions using the agent-council's 3-stage voting pipeline.
When to Use
Automatically invoke this skill when facing:
- Architectural design decisions that affect system structure
- Technology stack choices with multiple valid options
- Implementation approach evaluation for complex features
- Design pattern selection when trade-offs are unclear
- Multi-step planning for large tasks
How It Works
The skill runs the agent-council CLI which:
- Sends your question to Claude, Codex, and Gemini
- Each model provides an independent response
- Models peer-rank each other's answers
- A chairman synthesizes the final recommendation
Usage
When you encounter a complex decision point:
- Formulate the question clearly
- Run:
agent-council "your question here" - Wait for the 3-stage pipeline to complete
- Use the synthesized answer to guide implementation
Example Questions
- "What's the best way to implement real-time updates: WebSockets, SSE, or polling?"
- "Should this service use SQL or NoSQL database?"
- "What caching strategy would work best for this API?"
- "How should we structure error handling in this module?"
Prerequisites
The agent-council CLI must be installed:
npm install -g agent-council
Benefits
- Reduced bias: Multiple AI models reduce single-model bias
- Diverse perspectives: Different models have different strengths
- Quality ranking: Peer evaluation identifies best answers
- Synthesized wisdom: Chairman combines best insights