---

name: design-council description: Invoke a multi-model council for diverse perspectives on design questions

Design Council Skill

Consults multiple AI models on a design question and synthesizes consensus/dissent.

When to Use

Use this skill when you want to:

• Get diverse perspectives on an important design decision
• Reduce blind spots in your thinking
• Validate an approach before committing
• Surface potential issues you might have missed

Input Format

CopyQuestion: [Your design question]
Context: [Relevant background and constraints]
Models: [Optional - which models to consult, default: GPT-4, Claude, Gemini]

Examples

Architecture Decision

CopyQuestion: Should we use an event-driven architecture or request-response for inter-service communication?
Context: Building a real-time trading platform with strict latency requirements under 100ms. Team has experience with both patterns. Current scale is 10k requests/second.

Technology Selection

CopyQuestion: Rust vs Go for our new CLI tool?
Context: Tool will be distributed to non-technical users, needs cross-platform support, handles large file processing. Team knows Go but wants to learn Rust.
Models: gpt-4o, claude-3-opus

What Happens

1. Your question is sent to the LangGraph council endpoint
2. Each model in the roster generates an independent response
3. Responses are normalized and cross-critiqued
4. A synthesis produces consensus and dissent points

Response Format

Copy## Consensus
- All models agree that...
- Strong alignment on...

## Dissent
- GPT-4 favors X because...
- Claude prefers Y due to...
- Gemini raises concern about...

## Synthesis
Given the context, the recommended approach is...
Key considerations: ...
Potential risks: ...

Implementation

Call the LangGraph API at: http://localhost:8100/graphs/council

Copycurl -X POST http://localhost:8100/graphs/council \
  -H "Content-Type: application/json" \
  -d '{
    "question": "Should we use microservices or monolith?",
    "context": "Early-stage startup, team of 3...",
    "model_roster": ["gpt-4", "claude-3-opus"]
  }'