Debate Simulator — Debate Simulation Full Pipeline
An agent team collaborates to perform topic analysis, pro/con argument construction, cross-examination, judge evaluation, and comprehensive report writing.
Execution Mode
Agent Team — 5 agents communicate directly via SendMessage and simulate an actual debate.
Agent Roster
| Agent |
File |
Role |
Type |
| topic-analyst |
.claude/agents/topic-analyst.md |
Topic analysis, issue structuring |
general-purpose |
| pro-debater |
.claude/agents/pro-debater.md |
Pro-side argument construction |
general-purpose |
| con-debater |
.claude/agents/con-debater.md |
Con-side argument construction |
general-purpose |
| judge |
.claude/agents/judge.md |
Argument evaluation, verdict rendering |
general-purpose |
| rapporteur |
.claude/agents/rapporteur.md |
Comprehensive report writing |
general-purpose |
Workflow
Phase 1: Preparation (Orchestrator performs directly)
- Extract from user input:
- Resolution: A clear resolution that can be debated pro/con
- Debate purpose (optional): Educational / decision-making / competition prep / analysis
- User's position (optional): Pro / Con / Neutral
- Debate format (optional): Parliamentary / Lincoln-Douglas / Open debate
- Existing materials (optional): Related reports, papers, articles
- Create a
_workspace/ directory at the project root
- Organize the input and save it to
_workspace/00_input.md
- Determine the execution mode based on the scope of the request
Phase 2: Team Assembly and Execution
| Order |
Task |
Agent |
Depends On |
Deliverable |
| 1 |
Topic analysis |
topic-analyst |
None |
_workspace/01_topic_analysis.md |
| 2a |
Pro argument construction |
pro-debater |
Task 1 |
_workspace/02_pro_arguments.md |
| 2b |
Con argument construction |
con-debater |
Task 1 |
_workspace/03_con_arguments.md |
| 3 |
Cross-examination |
pro + con |
Tasks 2a, 2b |
_workspace/04_cross_examination.md |
| 4 |
Judge evaluation |
judge |
Tasks 2a, 2b, 3 |
_workspace/05_judge_verdict.md |
| 5 |
Comprehensive report |
rapporteur |
All |
_workspace/06_final_report.md |
Tasks 2a (pro) and 2b (con) run in parallel. However, since the con side may reference pro arguments, initial arguments are constructed independently and cross-referenced during the rebuttal phase.
Inter-agent communication flow:
- topic-analyst completes -> sends pro-favorable materials to pro-debater; con-favorable materials to con-debater; evaluation criteria to judge
- pro-debater + con-debater complete -> cross-examination simulation: exchange questions and answers
- cross-examination completes -> judge evaluates all arguments + cross-examination
- judge completes -> rapporteur synthesizes everything into a comprehensive report
Phase 3: Cross-Examination Simulation
The orchestrator mediates cross-examination:
- Deliver pro-debater's cross-examination questions to con-debater
- Record con-debater's answers
- Deliver con-debater's cross-examination questions to pro-debater
- Record pro-debater's answers
- Save the results to
_workspace/04_cross_examination.md
Phase 4: Integration and Final Report
- Verify all files in
_workspace/
- Verify the comprehensive report's balance and completeness
- Report the final summary to the user
Task-Scale Modes
| User Request Pattern |
Execution Mode |
Agents Deployed |
| "Run a debate simulation", "full debate" |
Full Simulation |
All 5 agents |
| "Just analyze the topic", "structure the issues" |
Analysis Mode |
topic-analyst only |
| "Just build pro and con arguments" |
Argument Mode |
topic-analyst + pro + con |
| "Write pro arguments for this topic" |
Single-Side Mode |
topic-analyst + pro-debater |
| "Evaluate this debate" (existing materials) |
Evaluation Mode |
judge + rapporteur |
Data Transfer Protocol
| Strategy |
Method |
Purpose |
| File-based |
_workspace/ directory |
Store and share major deliverables |
| Message-based |
SendMessage |
Real-time cross-examination exchange, revision requests |
| Task-based |
TaskCreate/TaskUpdate |
Track progress |
Error Handling
| Error Type |
Strategy |
| Resolution unclear |
Ask user to clarify, or have topic-analyst reframe |
| Pro/con balance impossible |
Adjust premises to reframe the resolution |
| Background materials insufficient |
Apply analogous cases; note "limited data" |
| Web search failure |
Work from general knowledge; note in report |
| Agent failure |
Retry once -> proceed without that deliverable |
Test Scenarios
Normal Flow
Prompt: "Run a debate simulation on the topic: 'AI replacing human jobs is socially desirable'"
Expected Results:
- Topic analysis: 3-4 structured issues, background materials, debate format
- Pro: 3-5 core claims + evidence + cross-examination questions
- Con: Pro rebuttal + independent claims + alternative proposal
- Cross-examination: 3 rounds of Q&A per side
- Judge: 100-point evaluation + verdict + feedback
- Report: Balanced comprehensive report + insights
Existing File Flow
Prompt: "Run cross-examination and judge evaluation based on these debate materials" + pro/con argument files attached
Expected Results:
- Existing materials copied to
_workspace/02_pro_arguments.md and _workspace/03_con_arguments.md
- Skip topic-analyst, pro-debater, con-debater; deploy cross-examination + judge + rapporteur
Error Flow
Prompt: "Run a debate simulation; pick a topic for me"
Expected Results:
- Resolution unclear -> topic-analyst proposes 3 current-event-based resolutions for selection
- After selection, proceed with full simulation
Agent Extension Skills
Extension skills that enhance each agent's domain expertise:
| Agent |
Extension Skill |
Role |
| pro-debater, con-debater |
argumentation-framework |
Toulmin argument model, evidence pyramid, 5-Type rebuttal strategies, cross-examination design |
| judge, rapporteur |
logical-fallacy-detector |
4-category logical fallacy classification, fallacy penalty standards, argument soundness rubric |
1---2name: debate-simulator3description: A debate simulation full pipeline. An agent team collaborates to perform topic analysis, pro argument construction, con argument construction, cross-examination, judge evaluation, and comprehensive report writing. Use this skill for requests like 'run a debate simulation', 'build pro and con arguments', 'debate', 'debate preparation', 'argument analysis', 'pro vs con', 'debate topic analysis', 'cross-examination practice', 'debate competition prep', and other debate-related needs. Existing arguments or debate materials can augment the relevant phase. However, live voice debate facilitation and audience voting system operation are outside the scope of this skill.4---56# Debate Simulator — Debate Simulation Full Pipeline78An agent team collaborates to perform topic analysis, pro/con argument construction, cross-examination, judge evaluation, and comprehensive report writing.910## Execution Mode1112**Agent Team** — 5 agents communicate directly via SendMessage and simulate an actual debate.1314## Agent Roster1516| Agent | File | Role | Type |17|-------|------|------|------|18| topic-analyst | `.claude/agents/topic-analyst.md` | Topic analysis, issue structuring | general-purpose |19| pro-debater | `.claude/agents/pro-debater.md` | Pro-side argument construction | general-purpose |20| con-debater | `.claude/agents/con-debater.md` | Con-side argument construction | general-purpose |21| judge | `.claude/agents/judge.md` | Argument evaluation, verdict rendering | general-purpose |22| rapporteur | `.claude/agents/rapporteur.md` | Comprehensive report writing | general-purpose |2324## Workflow2526### Phase 1: Preparation (Orchestrator performs directly)27281. Extract from user input:29 - **Resolution**: A clear resolution that can be debated pro/con30 - **Debate purpose** (optional): Educational / decision-making / competition prep / analysis31 - **User's position** (optional): Pro / Con / Neutral32 - **Debate format** (optional): Parliamentary / Lincoln-Douglas / Open debate33 - **Existing materials** (optional): Related reports, papers, articles342. Create a `_workspace/` directory at the project root353. Organize the input and save it to `_workspace/00_input.md`364. Determine the **execution mode** based on the scope of the request3738### Phase 2: Team Assembly and Execution3940| Order | Task | Agent | Depends On | Deliverable |41|-------|------|-------|-----------|-------------|42| 1 | Topic analysis | topic-analyst | None | `_workspace/01_topic_analysis.md` |43| 2a | Pro argument construction | pro-debater | Task 1 | `_workspace/02_pro_arguments.md` |44| 2b | Con argument construction | con-debater | Task 1 | `_workspace/03_con_arguments.md` |45| 3 | Cross-examination | pro + con | Tasks 2a, 2b | `_workspace/04_cross_examination.md` |46| 4 | Judge evaluation | judge | Tasks 2a, 2b, 3 | `_workspace/05_judge_verdict.md` |47| 5 | Comprehensive report | rapporteur | All | `_workspace/06_final_report.md` |4849Tasks 2a (pro) and 2b (con) run **in parallel**. However, since the con side may reference pro arguments, initial arguments are constructed independently and cross-referenced during the rebuttal phase.5051**Inter-agent communication flow:**52- topic-analyst completes -> sends pro-favorable materials to pro-debater; con-favorable materials to con-debater; evaluation criteria to judge53- pro-debater + con-debater complete -> cross-examination simulation: exchange questions and answers54- cross-examination completes -> judge evaluates all arguments + cross-examination55- judge completes -> rapporteur synthesizes everything into a comprehensive report5657### Phase 3: Cross-Examination Simulation5859The orchestrator mediates cross-examination:601. Deliver pro-debater's cross-examination questions to con-debater612. Record con-debater's answers623. Deliver con-debater's cross-examination questions to pro-debater634. Record pro-debater's answers645. Save the results to `_workspace/04_cross_examination.md`6566### Phase 4: Integration and Final Report67681. Verify all files in `_workspace/`692. Verify the comprehensive report's balance and completeness703. Report the final summary to the user7172## Task-Scale Modes7374| User Request Pattern | Execution Mode | Agents Deployed |75|---------------------|---------------|----------------|76| "Run a debate simulation", "full debate" | **Full Simulation** | All 5 agents |77| "Just analyze the topic", "structure the issues" | **Analysis Mode** | topic-analyst only |78| "Just build pro and con arguments" | **Argument Mode** | topic-analyst + pro + con |79| "Write pro arguments for this topic" | **Single-Side Mode** | topic-analyst + pro-debater |80| "Evaluate this debate" (existing materials) | **Evaluation Mode** | judge + rapporteur |8182## Data Transfer Protocol8384| Strategy | Method | Purpose |85|----------|--------|---------|86| File-based | `_workspace/` directory | Store and share major deliverables |87| Message-based | SendMessage | Real-time cross-examination exchange, revision requests |88| Task-based | TaskCreate/TaskUpdate | Track progress |8990## Error Handling9192| Error Type | Strategy |93|-----------|----------|94| Resolution unclear | Ask user to clarify, or have topic-analyst reframe |95| Pro/con balance impossible | Adjust premises to reframe the resolution |96| Background materials insufficient | Apply analogous cases; note "limited data" |97| Web search failure | Work from general knowledge; note in report |98| Agent failure | Retry once -> proceed without that deliverable |99100## Test Scenarios101102### Normal Flow103**Prompt**: "Run a debate simulation on the topic: 'AI replacing human jobs is socially desirable'"104**Expected Results**:105- Topic analysis: 3-4 structured issues, background materials, debate format106- Pro: 3-5 core claims + evidence + cross-examination questions107- Con: Pro rebuttal + independent claims + alternative proposal108- Cross-examination: 3 rounds of Q&A per side109- Judge: 100-point evaluation + verdict + feedback110- Report: Balanced comprehensive report + insights111112### Existing File Flow113**Prompt**: "Run cross-examination and judge evaluation based on these debate materials" + pro/con argument files attached114**Expected Results**:115- Existing materials copied to `_workspace/02_pro_arguments.md` and `_workspace/03_con_arguments.md`116- Skip topic-analyst, pro-debater, con-debater; deploy cross-examination + judge + rapporteur117118### Error Flow119**Prompt**: "Run a debate simulation; pick a topic for me"120**Expected Results**:121- Resolution unclear -> topic-analyst proposes 3 current-event-based resolutions for selection122- After selection, proceed with full simulation123124## Agent Extension Skills125126Extension skills that enhance each agent's domain expertise:127128| Agent | Extension Skill | Role |129|-------|----------------|------|130| pro-debater, con-debater | `argumentation-framework` | Toulmin argument model, evidence pyramid, 5-Type rebuttal strategies, cross-examination design |131| judge, rapporteur | `logical-fallacy-detector` | 4-category logical fallacy classification, fallacy penalty standards, argument soundness rubric |
revfactory