LLM Council

Run any question, idea, or decision through a council of 5 AI advisors who independently analyze it, peer-review each other anonymously, and synthesize a final verdict.

Based on Andrej Karpathy’s LLM Council methodology.

How it Works

1. Frame — Your question gets enriched with workspace context
2. Convene — 5 advisors analyze independently in parallel (Contrarian, First Principles, Expansionist, Outsider, Executor)
3. Peer Review — Advisors review each other’s work anonymously
4. Synthesize — A chairman produces the final verdict with agreements, clashes, blind spots, and a clear recommendation
5. Report — Visual HTML report + full markdown transcript

The Five Advisors

Triggers

• council this
• run the council
• war room this
• pressure-test this
• stress-test this
• debate this

Install

Copy SKILL.md to your Claude Code skills folder:

# Project-level
cp SKILL.md .claude/skills/llm-council/SKILL.md

# Global (all projects)
cp SKILL.md ~/.claude/skills/llm-council/SKILL.md

Or for OpenClaw users, place in your workspace skills/ directory.

Output

Every council session produces:

• council-report-[timestamp].html — Visual report for scanning
• council-transcript-[timestamp].md — Full transcript for reference

When to Use

✅ Decisions with real stakes and trade-offs ✅ Multiple viable options where being wrong is expensive ✅ Ideas that need pressure-testing from different angles

❌ Factual lookups with one right answer ❌ Creation tasks (write me a tweet) ❌ Processing tasks (summarize this)

Credits

• Methodology: Andrej Karpathy
• Skill adaptation: Bruno Okamoto
• Source article: Google Doc

License

MIT