council-brief
par jeadland
Web & Frontend DevelopmentRésumé
Unified LLM Council skill — install, query, and manage the multi-model consensus app.
Description
Ce que fait ce skill
LLM Council sends a question to multiple language models, has them critique each other's responses, and produces a synthesized "chairman" answer. This skill installs, queries, and manages the LLM Council app through a single command. Two modes are available: a headless CLI for quick answers and a full web UI for exploring model disagreements.
Instead of querying models one by one and manually comparing answers, LLM Council runs the deliberation and synthesis automatically.
Quand l'utiliser
- Comparing how different LLMs approach a tech stack decision
- Getting a consensus take on a financial or investment question
- Seeing where models disagree before making a product call
- Running a quick multi-model query from the terminal without a browser
- Hosting a local multi-model debate session for research or writing
Exemple de workflow
User asks: /council-brief "Is Python or Go better for microservices?"
- 1
Routes the question to ask-council.sh
- 2
Sends the question to the LLM Council backend on port 8001
- 3
Waits for Stage 1: all configured models submit responses
- 4
Waits for Stage 2: models rank and critique each other's answers
- 5
Returns Stage 3: the chairman's final synthesized answer
A single synthesized answer representing the council's consensus, delivered in the terminal after 30-60 seconds
Prérequis
- OPENROUTER_API_KEY environment variable (or OpenClaw gateway configured)
- uv (Python package manager) installed
Installation
clawhub install jeadland/council-briefExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.