Quorum
14-member multi-AI deliberation with structured ethical analysis
Overview
Quorum gathers opinions from 14 AI participants across a 3-stage deliberation process:
- Technical Council (7 members) - Diverse cognitive perspectives on the problem
- Ethical Sub-Chairs (6 members) - Analysis through ethical frameworks
- Chairman (1) - Final synthesis with transparency section
Architecture
THE CHAIRMAN (Opus 4.5) - Final Arbiter
│
├── Stage 1: TECHNICAL COUNCIL (7 experts)
│ ├── The Architect (GPT-5.2) - Strategic systems thinking
│ ├── The Scholar (Codex-Max) - Deep technical analysis
│ ├── The Sprinter (Codex-Mini) - Pragmatic speed
│ ├── The Diplomat (Sonnet 4) - Balanced human factors
│ ├── The Monk (Haiku 4.5) - First principles
│ ├── The Oracle (Gemini Pro) - Research synthesis
│ └── The Scout (Gemini Flash) - Pattern recognition
│
├── Stage 2: ETHICAL SUB-CHAIRS (6 frameworks)
│ ├── The Utilitarian - Consequentialism (outcomes)
│ ├── The Kantian - Deontology (duties)
│ ├── The Aristotelian - Virtue Ethics (character)
│ ├── The Pragmatist - Pragmatism (what works)
│ ├── The Guardian - Care Ethics (harm prevention)
│ └── The Machiavelli - Realpolitik (power dynamics)
│
└── Stage 3: CHAIRMAN SYNTHESIS
└── Verdict + Dissenting Views + "What Might Be Missed"
Trigger Conditions
This skill activates when:
- "Summon the council"
- "Get a quorum on this"
- "What do other AIs think?"
- "Multiple perspectives please"
- "quorum"
Usage
Direct Execution
~/.claude/skills/quorum/scripts/quorum.sh "Your question here"
CLI Options
| Option | Description |
|--------|-------------|
| -h, --help | Show usage information |
| -v, --verbose | Show detailed error output |
| -n, --dry-run | Preview without running |
| -b, --no-blind | Show real model names to chairman |
| -o, --output-dir DIR | Save reports to directory |
Output Files
Generates three markdown files:
quorum-minutes_TIMESTAMP.md- Full deliberationquorum-ethics_TIMESTAMP.md- Ethical synthesisquorum-chairman-report_TIMESTAMP.md- Final verdict
Chairman's Report Structure
- Battle Lines - Points of disagreement
- Ethical Spectrum - Framework analysis
- Pressure Test - Challenging arguments
- Uncomfortable Truths - What's avoided
- The Verdict - Recommendation
- Dissenting Views - Minority positions
- What Might Be Missed - Transparency
- Confidence - HIGH/MEDIUM/LOW
- Open Items - Remaining questions
Configuration
Edit quorum.config.yaml to customize members, commands, and settings.
Requirements
CLI tools must be installed:
codex- OpenAI Codex CLIgemini- Google Gemini CLIclaude- Anthropic Claude CLI
Cost
~14 API calls per session. With subscription plans, no additional costs.
微信扫一扫