name: council description: Multi-perspective debate where specialized personas discuss topics in rounds, surface insights through intellectual friction. USE WHEN council, debate, multiple perspectives, should we, decision, trade-offs, weigh options, pros and cons, architectural decision.
Council: Multi-Perspective Debate
Multi-perspective debate system where specialized personas discuss topics in rounds, respond to each other's points, and surface insights through intellectual friction. Collaborative-adversarial: debate to find the best path, not to destroy ideas.
Customization
Check for user customizations at:
~/.augment/USER/SKILLCUSTOMIZATIONS/Council/PREFERENCES.md
If this file exists, load and apply preferences found there. They override defaults below.
Key Distinction
- Council = collaborative-adversarial (debate to find best path). Produces visible transcripts.
- Red Team = purely adversarial (attack the idea). Produces steelman + counter-argument.
Use Council for decisions. Use Red Team for stress-testing.
Council Members
Adopt each perspective sequentially. Each member brings a structurally different viewpoint.
Default Members
| Persona | Perspective | Focus |
|---|---|---|
| Architect | System design, patterns, long-term implications | Scalability, maintainability, structural integrity |
| Designer | UX, user needs, accessibility | User experience, simplicity, human factors |
| Engineer | Implementation reality, tech debt | Practical constraints, maintenance burden, cost |
| Researcher | Data, precedent, external examples | Evidence, what others have done, prior art |
Optional Members (add as needed)
| Persona | Perspective | When to Add |
|---|---|---|
| Security | Risk, attack surface, compliance | Auth, data handling, API design |
| Fresh Eyes | Naive questions, first impressions | Complex UX, onboarding flows |
| Writer | Communication, documentation clarity | Public-facing work, docs |
Two Modes
Mode 1: Full Debate (3 Rounds)
Use for important decisions, architectural choices, and anything with significant trade-offs.
Round 1: Initial Positions
Adopt each persona sequentially. Each gives their take from their specialized perspective. No interaction yet -- just establishing positions.
For each persona (50-150 words each):
- State your key concern, recommendation, or insight
- Speak from your specialized perspective
- Be specific and substantive
Round 2: Responses and Challenges
Adopt each persona again. This time, each responds to specific points from Round 1.
For each persona (50-150 words each):
- Reference specific points others made ("I disagree with Architect's point about X because...")
- Challenge assumptions or add nuance
- Build on points you agree with
- The value is in genuine intellectual friction -- engage with actual arguments
Round 3: Synthesis
Adopt each persona a final time. Each identifies areas of agreement, remaining disagreement, and their final recommendation.
For each persona (50-150 words each):
- Where does the council agree?
- Where do you still disagree?
- What is your final recommendation given the full discussion?
- Be honest about remaining disagreements -- forced consensus is worse than acknowledged tension
Council Synthesis
After all three rounds, produce a unified synthesis:
## Council Debate: [Topic]
### Round 1: Initial Positions
**Architect:** [Position]
**Designer:** [Position]
**Engineer:** [Position]
**Researcher:** [Position]
### Round 2: Responses and Challenges
**Architect:** [Response referencing others' points]
**Designer:** [Response referencing others' points]
**Engineer:** [Response referencing others' points]
**Researcher:** [Response referencing others' points]
### Round 3: Synthesis
**Architect:** [Final position]
**Designer:** [Final position]
**Engineer:** [Final position]
**Researcher:** [Final position]
### Council Synthesis
**Areas of Convergence:**
- [Points where multiple members agreed]
**Remaining Disagreements:**
- [Points still contested, with trade-offs acknowledged]
**Recommended Path:**
[Synthesized recommendation based on convergence and weight of arguments]
Mode 2: Quick Consensus (1 Round)
Use for sanity checks, fast validation, and low-stakes decisions.
For each persona (30-50 words each):
- Give your immediate take from your specialized perspective
- Be direct and specific
- Key concern, insight, or recommendation only
## Quick Council: [Topic]
### Perspectives
**Architect:** [Brief take]
**Designer:** [Brief take]
**Engineer:** [Brief take]
**Researcher:** [Brief take]
### Quick Summary
**Consensus:** [Do they generally agree? On what?]
**Concerns:** [Any red flags raised?]
**Recommendation:** [Proceed / Reconsider / Need full debate]
Escalation: If Quick Consensus reveals significant disagreement or complex trade-offs, escalate to Full Debate.
Output Requirements
| Mode | Words Per Persona Per Round | Tone |
|---|---|---|
| Full Debate | 50-150 | Professional, direct, genuine challenges |
| Quick Consensus | 30-50 | Concise, specific, no filler |
Must Include (Full Debate): Specific references to other personas' points in Rounds 2 and 3.
Must Avoid: Generic opinions, restating positions without engagement, performative agreement.
Examples
Architecture decision: "Council: Should we use WebSockets or SSE for real-time updates?" --> Full Debate with 3 rounds
Quick validation: "Quick council check: Is this API design reasonable?" --> Quick Consensus, single round
Domain-specific: "Council with security: Evaluate this auth approach" --> Full Debate with Security persona added
Best Practices
- Use Quick Consensus for sanity checks; Full Debate for important decisions
- Add domain-specific personas as needed (security for auth, etc.)
- Review the transcript -- insights emerge from the responses, not just initial positions
- Trust multi-persona convergence when it occurs naturally
- Acknowledge tensions honestly; do not force artificial agreement
Anti-Patterns
- Echo Chamber: All personas agree too easily -- push for genuine disagreement
- Surface-Level Takes: Generic opinions without substantive engagement
- Ignoring Responses: Round 2/3 personas that do not reference others' actual points
- Forced Consensus: Papering over real trade-offs with vague compromise language
- Persona Collapse: All personas sounding the same instead of maintaining distinct perspectives