Agent skill

pal-consensus

Multi-model consensus building through structured debate using PAL MCP. Use for complex decisions, architectural choices, technology evaluations, or when you need multiple perspectives. Triggers on requests for second opinions, debates, or consensus building.

View SKILL.md on GitHub Repository
Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/development/pal-consensus-estiens-metacognitive-chat

SKILL.md

PAL Consensus - Multi-Model Debate

Build consensus through systematic analysis and structured debate across multiple AI models.

When to Use

  • Complex architectural decisions
  • Technology selection
  • Feature design trade-offs
  • Risk assessment
  • When you need multiple perspectives
  • Validating important decisions

Quick Start

python
# Step 1: State the proposal and do your analysis
result = mcp__pal__consensus(
    step="Evaluate: Should we migrate from monolith to microservices?",
    step_number=1,
    total_steps=4,  # Your analysis + 2 models + synthesis
    next_step_required=True,
    findings="My initial analysis: Consider scale, team size, complexity...",
    models=[
        {"model": "openai/gpt-5", "stance": "for"},
        {"model": "deepseek/deepseek-v3.2", "stance": "against"}
    ]
)

# Step 2+: Process each model's response
result = mcp__pal__consensus(
    step="Recording pro-microservices perspective",
    step_number=2,
    total_steps=4,
    next_step_required=True,
    findings="GPT-5 argues: scalability benefits, team autonomy...",
    continuation_id=result["continuation_id"]
)

Stance Types

Stance Description
for Advocate for the proposal
against Argue against the proposal
neutral Objective analysis without position

Required Parameters

Parameter Type Description
step string Proposal (step 1) or notes (step 2+)
step_number int Current step
total_steps int Models consulted + 2 (analysis + synthesis)
next_step_required bool More consultations needed?
findings string Your analysis or model response summary

Optional Parameters

Parameter Type Description
models list Models to consult with stances
current_model_index int Next model to consult
model_responses list Internal log of responses
relevant_files list Supporting files
continuation_id string Continue session

Model Configuration

python
models=[
    {
        "model": "openai/gpt-5",
        "stance": "for",
        "stance_prompt": "Focus on scalability and maintainability benefits"
    },
    {
        "model": "deepseek/deepseek-v3.2",
        "stance": "against",
        "stance_prompt": "Consider operational complexity and team capacity"
    },
    {
        "model": "google/gemini-3-flash-preview",
        "stance": "neutral"
    }
]

Available Models

Top models for consensus (all score 100):

  • openai/gpt-5 - Strong reasoning
  • deepseek/deepseek-v3.2 - Thinking-enabled
  • google/gemini-3-flash-preview - 1M context
  • x-ai/grok-4.1 - 2M context
  • bytedance-seed/seed-1.6 - Thinking-enabled

Workflow Pattern

Step 1: State proposal + your independent analysis
        ↓
Step 2: First model responds (for/against/neutral)
        ↓
Step 3: Second model responds
        ↓
Step N: Synthesize all perspectives

Example: Technology Decision

python
# Debate: GraphQL vs REST
mcp__pal__consensus(
    step="Evaluate: Should we use GraphQL instead of REST for our new API?",
    step_number=1,
    total_steps=5,
    next_step_required=True,
    findings="""
    Initial analysis:
    - Current team: 5 backend devs, familiar with REST
    - Use case: Mobile app with varying data needs
    - Timeline: 3 months to launch
    - Consider: Learning curve, tooling, performance
    """,
    models=[
        {"model": "openai/gpt-5", "stance": "for"},
        {"model": "deepseek/deepseek-v3.2", "stance": "against"},
        {"model": "google/gemini-3-flash-preview", "stance": "neutral"}
    ],
    relevant_files=[
        "/docs/api-requirements.md",
        "/app/api/current_endpoints.py"
    ]
)

Best Practices

  1. Frame proposals clearly - Specific, evaluable statements
  2. Provide context - Constraints, requirements, history
  3. Use diverse models - Different strengths and perspectives
  4. Balance stances - Include for, against, and neutral
  5. Document synthesis - Capture key insights from all perspectives

Didn't find tool you were looking for?

Be as detailed as possible for better results