Debate
ProductivitydebateDebate is an Agent Skill for productivity published in the majiayu000/claude-skill-registry repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Use when evaluating competing options or stress-testing a proposal, typically after brainstorming. Supports quick, thorough, adversarial, and collaborative styles.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/majiayu000/claude-skill-registry/blob/HEAD/skills/orchestration/debate-cmtkdot-arc/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/debate, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Debate
Run structured debate rounds with assigned roles to evaluate competing options or stress-test a proposal. Use this in the Full tier after brainstorming, or any time a decision needs rigorous evaluation.
Debate Styles
Quick
Two rounds maximum. One proponent, one opponent. Best for binary decisions with clear tradeoffs. Time-boxed and focused on the top 2-3 differentiating factors.
Thorough
Three to five rounds. Multiple roles (proponent, opponent, domain expert, end-user advocate). Covers technical feasibility, user impact, maintenance cost, and risk. Use when the decision has significant long-term consequences.
Adversarial
Dedicated red-team approach. The opponent actively tries to break the proposal by finding edge cases, failure modes, security issues, and scalability limits. The proponent must defend or concede each point. Use for security-sensitive or high-risk architectural decisions.
Collaborative
All participants work toward the best synthesis rather than defending fixed positions. Each round builds on the previous one. Participants may change positions as new information surfaces. Use when the goal is refinement rather than selection.
Round Structure
Each round follows this sequence:
- Opening statement — Each role presents their position with supporting evidence.
- Challenge — Opposing roles identify weaknesses, missing considerations, or counterexamples.
- Response — Challenged roles address the points raised, conceding where appropriate.
- Synthesis — Summarize what was established, what remains contested, and what new information emerged.
Output Format
## Debate: [Topic]
### Style: [quick|thorough|adversarial|collaborative]
### Round N
**[Role]**: [position and evidence]
**[Role]**: [challenge or response]
**Synthesis**: [what was established]
### Verdict
**Decision**: [selected option or refined proposal]
**Confidence**: [high|medium|low]
**Key factors**: [top 3 deciding factors]
**Dissent**: [any unresolved objections worth noting]
Claude-Only Mode
When external providers are unavailable, simulate multiple perspectives by:
- Explicitly labeling each role before writing its argument.
- Genuinely steelmanning each position — do not create strawmen.
- Tracking concessions and adjustments across rounds.
- Producing a final synthesis that reflects the actual arguments made, not a predetermined conclusion.
When to Use Debate vs. Brainstorm
- Brainstorm generates options. Use when the solution space is unexplored.
- Debate evaluates options. Use when 2-4 clear alternatives exist and a decision is needed.
A common Full-tier sequence: brainstorm to generate options, then debate to select among them.