What is the difference between workflow-debate and workflow-brainstorm?

Brainstorming is for divergent idea generation. Debate is for convergent evaluation; it uses adversarial roles to find flaws and risks in an existing plan.

How do I trigger a debate in Claude Code?

You can use the /council command or phrases like 'debate this', 'stress-test my plan', or 'evaluate this idea' followed by your proposal.

What kind of personas participate in the council?

Depending on the configuration, the council includes a Visionary (opportunities), Critic (risks), Pragmatist (feasibility), Innovator (alternatives), Ethicist, and specialized Domain Experts.

Do I need special settings to use this skill?

Standard mode requires `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` in your settings.json. If you don't have this enabled, you can use the `-e` flag for Economy Mode.

Can I save the results of the debate?

Yes, by using the `-s` flag, the skill will save a full transcript of the deliberation to the `.claude/output/council/` directory.

AI Council Workflow Debate

Name: AI Council Workflow Debate
Author: flosrn

byflosrn

0•

生産性とワークフロー

Orchestrates structured adversarial debates between specialized AI personas to rigorously stress-test ideas, architectures, and strategic plans.

The workflow-debate skill implements a high-stakes AI Council Framework designed to move beyond simple brainstorming. By deploying a team of specialized agents—including Visionary, Critic, Pragmatist, and Ethicist roles—it facilitates a multi-round cross-examination of any proposal. This process surfaces hidden risks, identifies blind spots, and provides a structured verdict (Proceed, Reconsider, or Do Not Proceed) backed by a risk assessment matrix. Based on research showing that multi-perspective deliberation significantly improves decision accuracy, this skill is essential for evaluating complex technical migrations, security strategies, or product roadmaps.

主な機能

01Multi-agent adversarial deliberation with cross-examination rounds

02Specialized persona roles: Visionary, Critic, Pragmatist, Innovator, Ethicist, and Domain Expert

03Structured output including a risk assessment matrix and steelmanned counter-arguments

04Standard Mode using experimental Agent Teams or Economy Mode for sequential execution

050 GitHub stars

06Configurable intensity with adjustable councillor counts (3-6) and debate rounds (1-4)

ユースケース

01Stress-testing security protocols and data handling policies before implementation

02Validating strategic business decisions through multiple critical lenses to avoid groupthink

03Evaluating major architectural shifts like migrating from monoliths to microservices

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add flosrn/.claude workflow-debate

For use in Claude.ai and ChatGPT