llm-council

【开发与工程】Convene a council of independent agents that answer a hard question separately, blind-review and rank each other's answers, then synthesize one vetted answer. Fights sycophancy and one-angle reasoning — use when the user wants a real answer instead of agreement. Two modes — default (5 Claude sub-agents) and cross-vendor (real GPT/Gemini/Claude/Grok v…

独立 GitHub 快照 · 自动分类 · 以原作者文档为准