Can I run this skill with only the Claude CLI installed?

Yes, the skill supports graceful degradation. While 2 or more models are recommended for peer review, it can operate in a single-model mode for testing if only Claude is available.

How does the Chairman ensure an unbiased synthesis?

The Chairman agent operates in an isolated context with restricted tools (Read/Write only), ensuring its synthesis is based strictly on the deliberated files rather than its own external knowledge.

How does the three-phase protocol work?

It begins with Phase 1 (Parallel Opinion Collection), proceeds to Phase 2 (Peer Review where models critique each other), and concludes with Phase 3 (Chairman Synthesis) to produce a unified final report.

Where are the deliberation results stored?

All intermediate opinions, peer reviews, and the final markdown report are stored in a dedicated local '.council' directory, allowing for full auditability of the decision-making process.

Which LLMs are supported by the Council Orchestrator?

The skill supports Claude (required), OpenAI Codex, and Google Gemini. It can run in single-model mode or full consensus mode depending on which CLIs are installed and configured.

Council Orchestrator

Name: Council Orchestrator
Author: xrf9268-hue

byxrf9268-hue

•

Seguridad y Pruebas

Orchestrates multi-model LLM consensus through a three-phase deliberation protocol for collaborative decision-making and high-integrity code reviews.

Council Orchestrator is a specialized skill that leverages the collective intelligence of multiple large language models—including Claude, OpenAI Codex, and Google Gemini—to reach a verified consensus on complex technical tasks. By executing a structured three-phase protocol consisting of parallel opinion collection, cross-examination peer review, and chairman synthesis, it significantly reduces individual model biases and hallucinations. This skill is ideal for high-stakes scenarios such as architectural planning, security audits, and complex debugging where multiple expert perspectives are required to ensure the most robust solution.

Características Principales

01Automated cross-examination where models anonymously critique peer responses

02Graceful degradation and flexible quorum requirements based on available CLIs

037 GitHub stars

04Multi-model orchestration supporting Claude, OpenAI Codex, and Google Gemini

05Three-phase consensus protocol: Opinion, Peer Review, and Synthesis

06Isolated Chairman synthesis for unbiased, evidence-based final reporting

Casos de Uso

01Collaborative problem-solving to minimize AI hallucinations and errors

02Complex architectural decision-making and pattern validation

03High-stakes code reviews requiring diverse technical perspectives

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add xrf9268-hue/llm-council-plugin council-orchestrator

For use in Claude.ai and ChatGPT

Download Skill