LaunchKit · 2026
Back to Skills

Multi-agent debate council — spawns 3 specialized sub-agents in parallel (Scholar, Engineer, Muse) for Round 1.

2
534 downloads
by @robbyczgw-cla

Setup & Installation

openclaw skills install @robbyczgw-cla/roundtable

Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:

npx clawhub install roundtable

Version History

v0.4.1Mar 3, 2026 - Version note by skill developer:

Add package.json (ClawHub fix), version badges, improved docs

What This Skill Does

Spawns three specialized sub-agents (Scholar, Engineer, Muse) in parallel to analyze a question from research, logic, and creative angles. A coordinator agent synthesizes their outputs, with an optional second round where agents cross-examine each other's findings and update their positions. Models per role are configurable via flags or a saved config file.

Running parallel specialist agents with an optional cross-examination round surfaces contradictions and blind spots that a single-agent response typically misses.

When to use it

  • Comparing database or infrastructure options for a new project
  • Getting a code review from research, logic, and readability angles at once
  • Evaluating a financial or investment decision with structured debate
  • Stress-testing architectural trade-offs before committing to an approach
  • Diagnosing a complex problem that has no obvious single correct answer

Example Workflow

Here's how your AI assistant might use this skill in practice.

INPUT

User asks: /roundtable Should I use PostgreSQL or MongoDB for a new SaaS app?

AGENT
  1. 1Captain parses the command and decomposes the question into role-specific sub-tasks
  2. 2Scholar, Engineer, and Muse sub-agents are spawned in parallel with focused prompts and security wrappers
  3. 3Each agent returns structured findings with confidence ratings and dissent notes
  4. 4Captain summarizes Round 1, then spawns three Round 2 agents to challenge each other's conclusions
  5. 5Captain synthesizes consensus, conflicts, and any changed positions into a final answer
OUTPUT

A structured council answer with confidence level, points of agreement, dissent rationale, and consolidated source citations