swarm-discussion

Orchestrate structured multi-expert discussions with tension maps and argument graphs.

1|Updated Jan 29, 2026
One-click install
npx skills add https://github.com/Ischca/swarm-discussion-skill --skill swarm-discussion
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: swarm-discussion
Source: https://github.com/Ischca/swarm-discussion-skill/tree/main/.claude/skills/swarm-discussion
Command: npx skills add https://github.com/Ischca/swarm-discussion-skill --skill swarm-discussion

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill orchestrates structured multi-expert discussions to surface blind spots, prevent echo chambers, and generate robust, debate-driven insights through team-based messaging and tension-aware roles.

Core Features & Use Cases

  • Team-based Architecture: Compose expert teams via a Teammate-like API to drive diverse perspectives.
  • Messaging-based Dialogue: Enable direct, threaded discussion among experts with traceable references.
  • Structured Disagreement Protocol: Design tension maps, position declarations, and steel-manning to avoid premature convergence.
  • Argument Graph: Track claims, rebuttals, and evidence across rounds for reproducible reasoning.
  • Quality Gates & Convergence: Automatic quality scoring to manage disagreements and track evolution.
  • Dynamic Expert Generation: Generate topic-specific experts to adapt to new topics; supports 2-4 dynamic agents.
  • Live Progress & History: Real-time updates and historian-logged round records for transparency.
  • Minority Reports: Preserve dissenting viewpoints even when outnumbered.
  • User Participation: Involve users in defining goals and constraints.

Quick Start

Example: To begin a swarm-discussion session, issue a command like: /swarm-discussion "Is there a better way to manage complex group decision-making?"

Frequently Asked Questions about swarm-discussion

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

FAQPage Schema
How do I facilitate multi-expert debates to prevent echo chambers in complex decisions?

Multi-expert debates prevent echo chambers by using structured disagreement protocols like tension maps and steel-manning to force diverse perspectives. This approach tracks claims and rebuttals in an argument graph, ensuring conflicting viewpoints are fully explored and documented before reaching convergence.

What is a tension map and how does it improve group decision-making?

A tension map is a structured disagreement tool that explicitly charts conflicting viewpoints and positions among experts. It improves group decision-making by preventing premature convergence, ensuring all dissenting perspectives are declared and evaluated before a final synthesis is generated.

How do I track arguments and evidence across multiple rounds of expert discussion?

You track arguments across discussion rounds by using an argument graph to map claims, rebuttals, and evidence. This is paired with live progress reporting and historian-logged round records to maintain transparent, reproducible reasoning throughout the debate.

Can I dynamically generate topic-specific experts for a multi-agent debate?

Yes, you can dynamically generate topic-specific experts to adapt the discussion to any new subject. This multi-agent setup supports two to four dynamic agents, automatically assigning tension-aware roles to drive diverse perspectives and structured dialogue.

What is the best way to preserve minority viewpoints during AI-driven group discussions?

The best way to preserve minority viewpoints is through structured disagreement protocols that generate minority reports. By implementing quality gates and position declarations, dissenting opinions are documented and maintained even when outnumbered by the majority consensus.

Are there limitations to using structured argument graphs for team-based dialogue?

Structured argument graphs require consistent quality scoring and explicit position declarations to function effectively. You should avoid this approach if your decision domain lacks conflicting viewpoints, as tension maps and steel-manning protocols add overhead when consensus is already present.