← ClaudeAtlas

llm-councillisted

Run high-stakes questions, architectural decisions, product dilemmas, or major tradeoffs through a council of 5 specialized AI advisors who independently analyze the issue, peer-review each other anonymously, and deliver an actionable synthesized verdict. Based on Karpathy's LLM Council methodology, fully optimized for Google Antigravity multi-agent orchestration. MANDATORY TRIGGERS: 'council this', 'run the council', 'war room this', 'pressure-test this', 'stress-test this', 'debate this'. STRONG TRIGGERS: 'should I X or Y', 'which option should I choose', 'is this the right move', 'validate this architecture', 'get multiple perspectives', 'I can't decide between'.
Pranav-Nexus/antigravity-skill-porter · ★ 5 · AI & Automation · score 74
Install: claude install-skill Pranav-Nexus/antigravity-skill-porter
# LLM Council (Google Antigravity Edition) When tackling high-stakes decisions, a single model response often produces a single, un-scrutinized perspective. The **LLM Council** solves this by dispatching the decision across **5 distinct cognitive lenses**, running a round of **blind peer-review**, and synthesizing a **decisive chairman verdict** with concrete next actions. Adapted from Andrej Karpathy's LLM Council methodology and recompiled natively for **Google Antigravity's parallel subagent architecture**. --- ## When to Run the Council The Council is designed for situations where being wrong is costly or irreversible: ### High-Value Scenarios - **Architectural & Tech Stack Decisions**: *"Should we migrate our auth from Firebase to Supabase or custom OAuth?"* - **Product & Monetization Strategy**: *"Should we release a self-serve tier or stay enterprise-only?"* - **Refactoring & Paradigm Shifts**: *"Should we rewrite the frontend to Next.js or keep the existing Vue SPA?"* - **Pivots & Strategic Risk**: *"Should we build in-house or integrate an off-the-shelf vendor?"* ### Not Suitable For - Factual lookups (*"What is the syntax for TypeScript utility types?"*) - Simple code generation tasks without major architectural tradeoffs. - Low-stakes trivial queries. --- ## The Five Council Advisors Each advisor operates from an uncompromising mental model to generate natural creative friction: | Advisor | Cognitive Lens | Core Focus | | :--- | :--- | :--- | | **1. The