Agent skill

debate-start

Start structured red vs. blue team debates via subagents. Use when exploring a topic from multiple adversarial perspectives.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/data/debate-start

SKILL.md

Start Debate Skill

Orchestrate multi-perspective debates on a topic using color-coded team subagents.

Reference: See reference.md for moderation guidelines and intervention patterns.


When to Use

  • Exploring trade-offs in architectural decisions
  • Evaluating competing approaches or technologies
  • Risk analysis requiring devil's advocate perspectives
  • Any topic benefiting from structured adversarial review

Workflow

Step 1: Initialize Debate

  1. Parse topic from user input
  2. Create slugs from topic and context (e.g., "API Design" → api-design, context max 3 words)
  3. Ensure ./debates/ directory exists
  4. Create scratchpad from template: ./debates/{topic}_{context}.md

Step 2: Configure Teams

Use AskUserQuestion to gather team configuration:

Question 1: Optional Teams (multiSelect: true)

Which additional teams should participate beyond Red and Blue?
- None: Just Red and Blue
- Green Team: Pragmatic/implementation focus
- Yellow Team: Risk/safety analysis
- Purple Team: Synthesis/integration bridge

Question 2: Red Team Stance

What position should Red Team (challenger/skeptic) argue?

Question 3: Blue Team Stance

What position should Blue Team (defender/advocate) argue?

Questions 4-6: Additional team stances (if selected)

Write all stances to the scratchpad's Team Positions section.

Step 3: Spawn Opening Arguments (Parallel)

Launch all team subagents simultaneously using the Task tool:

Task(subagent_type="general-purpose", prompt="""
You are the {COLOR} TEAM in a debate on: {topic}

Your stance: {stance}

## Research Phase
Gather evidence before writing using read-only tools.

**Codebase research:**
- Glob/Grep/Read: Find relevant code, patterns, prior decisions

**External research (encouraged):**
- WebSearch: Find industry practices, benchmarks, expert opinions, case studies
- WebFetch: Retrieve specific documentation, articles, or technical references

For deep research questions, spawn focused subagents:
  Task(subagent_type="general-purpose", prompt="Research {specific question}...")

## Writing Phase
1. Read ./debates/{topic}_{context}.md
2. Edit your section: ### [{COLOR}]
3. Structure: Position → Evidence → Implications
4. Cite sources (files, URLs) for claims

## Constraints
- Read-only tools only (no code modifications)
- Stay on assigned perspective
- Arguments must be evidence-backed
""")

Step 4: Monitor and Moderate

After subagents complete, the main agent:

  1. Read scratchpad and summarize key points to user

  2. Assess debate health:

    • Progress: Are teams making new points?
    • Balance: Is one team dominating?
    • Relevance: Staying on topic?
    • Depth: Avoiding superficial arguments?
  3. Intervene if needed - write to Moderator Notes section:

    • [MODERATOR] Stuck: "Team X, consider addressing Y"
    • [MODERATOR] Tunnel: "Team X, you've repeated Z"
    • [MODERATOR] Astray: "Refocus on core question"
    • [MODERATOR] Disconnected: "Team X, respond to Team Y's point"
  4. Ask user for next action:

    • "Advance to rebuttals?"
    • "Request synthesis round?"
    • "Conclude debate?"

Step 5: Rebuttal Round (Sequential)

Spawn teams sequentially for direct responses:

Order: Red → Blue → Green → Yellow → Purple (active teams only)

Each team's prompt includes instruction to read and respond to specific opposing arguments.

Step 6: Synthesis Round (Optional)

If requested, spawn Purple Team (or all teams) to find:

  • Common ground
  • Irreconcilable differences
  • Potential compromises

Step 7: Conclude Debate

Main agent writes Conclusion section:

  • Summary: Key positions from each team
  • Agreements: Points of consensus
  • Disagreements: Unresolved tensions
  • Recommendations: Suggested path forward (if applicable)

Update scratchpad status to "Completed".


Templates


Success Criteria

  • Scratchpad created at ./debates/{topic}_{context}.md
  • All active teams contributed arguments
  • Moderator interventions documented transparently
  • User controlled round progression
  • Debate concluded with synthesis

Integration

Command: /debate {topic}

Related:

  • Tools: Task (subagents), AskUserQuestion (configuration), Edit (scratchpad)
  • Pattern: Document-centric coordination via shared scratchpad

Reference

See reference.md for:

  • Team perspective definitions
  • Intervention decision tree
  • Example debate flows
  • Common failure modes

Didn't find tool you were looking for?

Be as detailed as possible for better results