skills/debate/SKILL.md
Start structured red vs. blue team debates via subagents. Use when exploring a topic from multiple adversarial perspectives.
npx skillsauth add srnnkls/tropos debateInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Orchestrate multi-perspective debates on a topic using color-coded team subagents.
Reference: See reference.md for moderation guidelines and intervention patterns.
api-design, context max 3 words)./debates/ directory exists./debates/{topic}_{context}.mdUse 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.
Launch all team subagents simultaneously using the Task tool:
Task(subagent_type="general", 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", 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
""")
After subagents complete, the main agent:
Read scratchpad and summarize key points to user
Assess debate health:
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"Ask user for next action:
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.
If requested, spawn Purple Team (or all teams) to find:
Main agent writes Conclusion section:
Update scratchpad status to "Completed".
./debates/{topic}_{context}.mdCommand: /debate {topic}
Related:
See reference.md for:
tools
External code-review harness (`peer` zsh tool): canonical model registry, idle-stall watchdog, and self-parallelising fan-out to codex/gemini. Use when dispatching external (non-Claude) reviewers from the review or implement pipelines — call `peer run`/`peer <harness>` instead of `codex exec`/`gcloud`+Vertex directly.
testing
Unified validation dispatcher. Auto-detects validation type from argument or presents selection menu. Routes to test, implement (verify), or hooks-test.
development
Test-driven development methodology (RED-GREEN-REFACTOR). Use when implementing features, fixing bugs, or changing behavior - write failing test first, then minimal code to pass.
development
Create new Claude Code skills following project patterns and best practices. Use when building new skills, extracting reusable capabilities, or converting commands to skills.