llm-council

Runs decisions through five AI advisor sub-agents with peer review and chairman synthesis.

Updated Jul 17, 2026
One-click install
npx skills add https://github.com/guneysol/agent-configs --skill llm-council-guneysol
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: llm-council
Source: https://github.com/guneysol/agent-configs/tree/main/agents/skills/llm-council
Command: npx skills add https://github.com/guneysol/agent-configs --skill llm-council-guneysol

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? A single AI answer gives you only one perspective, with no way to judge its quality. This Skill pressure-tests important decisions by running them through five independent advisors with different thinking styles, then having them peer-review each other before a chairman synthesizes a final verdict. ## Core Features & Use Cases - Five-Perspective Analysis: Spawns Contrarian, First Principles, Expansionist, Outsider, and Executor sub-agents in parallel, each arguing fully from its own angle. - Anonymous Peer Review: Advisor responses are anonymized and cross-reviewed to surface the strongest argument, biggest blind spots, and what everyone missed. - Chairman Verdict: A final synthesis delivers where the council agrees, where it clashes, a direct recommendation, and one concrete next step. - Use Case: You're deciding between launching a $97 workshop or a $497 course. The council stress-tests pricing, positioning, and execution risk from five angles, then gives you a clear recommendation instead of a hedge. ## Quick Start Ask the assistant to council this decision: should I hire a VA or build an automation first, given my current workload and budget?

Frequently Asked Questions about llm-council

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a decision through multiple AI perspectives?▼

Trigger the council with a phrase like 'council this' followed by your decision. The skill spawns five advisor sub-agents in parallel, each analyzing from a different thinking style, then synthesizes their output into a final verdict.

What kinds of questions work best with an LLM council?▼

Genuine decisions with real stakes and multiple options, such as pricing choices, pivots, or hiring tradeoffs. It is not suited for factual lookups, simple yes/no questions, or content creation tasks.

How does the peer review step improve the council output?▼

Advisor responses are anonymized as Response A through E, then five reviewer sub-agents identify the strongest response, the biggest blind spot, and what all responses missed. Anonymization prevents deference to particular thinking styles.

When should I not use the LLM council approach?▼

Skip it for questions with one right answer, casual 'should I' questions without meaningful tradeoffs, and pure creation or summarization tasks. The council adds value only when genuine uncertainty and multiple valid perspectives exist.

Does the council verdict always follow the majority of advisors?▼

No. The chairman can side with a dissenting advisor if that reasoning is strongest, and must explain why. The verdict always ends with a direct recommendation and one concrete first step, not a hedge.