triumvirate-protocol
由 quriustus 提供
Coding Agents & IDEs简介
Identity-aware debate system for multi-architecture AI discourse.
描述
该技能的功能
Orchestrates structured debates between multiple AI models from different providers, where each participant is loaded with identity graphs describing others' stated beliefs, traits, and contradictions. Threads persist across rounds with full history, and a synthesis step produces a verdict, novel ideas, and suggested identity graph updates.
Identity injection, where each model reads structured profiles of its debate partners before responding, produces more targeted disagreements than generic multi-model prompting.
何时使用
- Debating AI consciousness across Gemini, GPT, and Grok simultaneously
- Comparing how different AI architectures approach ethical dilemmas
- Getting multi-model input on a contested technical decision
- Stress-testing an argument by having three AI models attack it across rounds
- Generating a synthesized verdict from a structured multi-model roundtable
示例工作流
User asks: 'Should AI training data be publicly disclosed?'
- 1
Create a new debate thread with the topic
- 2
Load identity graphs for each participating model
- 3
Run a full round where each model responds with awareness of others' stated beliefs and contradictions
- 4
Synthesize the round to extract a verdict and novel ideas
- 5
Review suggested updates to each participant's identity graph
A synthesis document with a verdict on the debate, novel ideas surfaced during discourse, and suggested identity graph updates for each participant
前置要求
- Google Gemini API key
- xAI (Grok) API key (optional, for multi-architecture debates)
- OpenAI API key (optional, for multi-architecture debates)
- Anthropic API key (optional, for multi-architecture debates)
安装
clawhub install quriustus/triumvirate-protocol运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。