llm-councillisted

# LLM Council Ask one AI, get one answer. It might be great, it might be mid, and you have no way to tell because you only saw one perspective. The council fixes that. Five advisors analyze the question independently from fundamentally different angles, then peer-review each other's work blind and rank it, then a chairman synthesizes where they agree, where they clash, and what you should actually do. Adapted from [Andrej Karpathy's LLM Council](https://github.com/karpathy/llm-council). He dispatches a query to several different models from different labs, has them rank each other anonymously, and a chairman compiles the final answer. This runs the same three stages inside Claude Code, substituting **thinking lenses** for **different models** as the source of diversity — a deliberate choice, not a limitation. See "Fidelity notes" at the end for why, and for what the substitution costs. Run all four phases. Do not shortcut one. ## Portability Self-contained. Depends only on built-in Claude Code tools — the `Agent` tool with the built-in `general-purpose` agent, plus `Glob`, `Read`, and `Write` — and on `report-template.html` in this same folder, which is used only by the optional Phase 5. No external scripts, APIs, paid services, or required skills. Drop the folder into any `.claude/skills/` directory and it works. A brand-identity skill, if the installation happens to have one, is used to restyle the HTML report. It is optional; without it the bundled neutral template