A structured 6-advisor council for Claude Code. Fetches real data, debates your decision from six conflicting angles, peer-reviews itself anonymously, and produces a Chairman's verdict with kill criteria and follow-up tracking.
Walkthrough: coherencedaddy.com/tutorials/advisory-board — 11-slide visual tutorial.
/plugin marketplace add Coherence-Daddy/advisory-board
/plugin install advisory-board@advisory-board
/reload-plugins
Use advisory-board to council this:
I'm thinking of [decision]. Audience [profile]. Constraints [budget, time].
What's the best move?
The facilitator scans your workspace, fetches real-world data via a recon agent, frames the question, and dispatches six advisors in parallel: Contrarian, First Principles Thinker, Expansionist, Customer, Executor, Operator. Each has mandatory checks they must address. Their responses get anonymized and peer-reviewed (5 questions each, including a falsification condition). The Chairman synthesizes into a verdict with kill criteria, a falsification trigger, and a single first action.
Three artifacts get saved per session:
council-report-<ts>.html— visual briefing (auto-opens)council-transcript-<ts>.md— full record with anonymization map revealedcouncil-followups-<ts>.md— pre-action checklist + re-council triggers
- You've already counciled this topic recently and the follow-up is narrow.
- The question is about executing a prior recommendation (just answer it).
- The question is factual or mechanical, not a multi-angle decision.
A council run is ~12 LLM calls plus recon. Don't council trivial questions.
- SKILL.md — the full facilitator instructions
- autoresearch — sibling plugin for optimizing skill files
MIT