advisory-board

Orchestrate isolated multi-agent deliberation and adjudication for high-stakes decisions.

Updated Sep 14, 2026
One-click install
npx skills add https://github.com/Lawrence908/chiron --skill advisory-board-lawrence908
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: advisory-board
Source: https://github.com/Lawrence908/chiron/tree/main/plugins/meta/skills/advisory-board
Command: npx skills add https://github.com/Lawrence908/chiron --skill advisory-board-lawrence908

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? High-stakes decisions often get a single agreeable AI answer that hides real tradeoffs. This Skill convenes a council of agents with deliberately conflicting mandates, runs them in isolation so none can soften against another, then forces one adjudicator to resolve the conflict into a ranked, defensible recommendation. ## Core Features & Use Cases - Structured disagreement: Council members (Pessimist, Optimist, CFO, SRE, etc.) argue in isolation from identical dossiers, so genuine conflicts surface instead of being averaged away. - Weighted adjudication: Criteria weights are locked before any member runs, and the Oracle adjudicator must name the deciding factor and the strongest objection it overrules. - Persisted, re-testable runs: Every run is saved with frame, dossier, member arguments, and verdict, so a past decision can be re-tested under changed weights via the weight-sweep script. - Use Case: Deciding whether to relocate a family, migrate infrastructure, or change business strategy — convene the family, technical, or executive roster, get a ranked recommendation with named dissent, and publish a board-record artifact. ## Quick Start Convene an advisory board using the family roster to evaluate whether we should relocate to another city this year.

Frequently Asked Questions about advisory-board

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

FAQPage Schema
How do I run a multi-agent advisory board for a big decision?▼

Provide a problem frame with stakes and options, a context dossier with real numbers and named people, and 5-9 weighted criteria summing to 100. The skill runs council members in isolation by order group, then an adjudicator produces a ranked verdict with named dissent.

What rosters are available for the advisory board council?▼

Four bundled rosters exist: family (Pessimist, Optimist, Liberator, Oracle), executive (C-suite roles), technical (SRE, Security, Architect, Maintainer, Oracle), and homelab (technical plus Liberator). Custom rosters are added as new markdown files without editing the skill.

Can I re-test a past decision under different criteria weights?▼

Yes. Re-runs create a new timestamped directory recording the parent run, and the weight-sweep.js script re-adjudicates the same member outputs under multiple weight configurations. Only the Oracle re-runs per weighting; members argue once.

When should I not use an advisory board process?▼

Skip it for factual lookups, questions with one correct answer, and small reversible choices where deliberation costs more than the mistake. It is designed for consequential, hard-to-reverse decisions needing structured disagreement.

Why must council members never see each other's output?▼

Isolation is the core mechanism: if members see each other, they soften toward consensus and the conflicting mandates collapse. Separate agent contexts preserve genuine disagreement, which the adjudicator then resolves explicitly rather than averaging.