Judge - Council of Experts

Multi-model council for comprehensive analysis. Adapts to whichever API keys you have configured.

Usage

/judge [your question or code]

Instructions

When user invokes /judge [query]:

Step 0: Check Available Tools

Before calling any tool, check which tachibot-mcp tools are available in this session by looking at the loaded MCP tools. The pipeline adapts:

• Has Grok? Use grok_search + grok_reason
• Has Perplexity? Use perplexity_ask for search
• Has OpenAI? Use openai_reason as first judge
• Has Gemini? Use gemini_analyze_text as final judge
• Has Qwen? Use qwen_coder for code analysis
• Has Kimi? Use kimi_thinking for step-by-step
• Has OpenRouter? Add deepseek_reason (frontier reasoning) + glm_reason (agentic) for lab diversity

Minimum viable council: any 2 models from different providers. If only 1 model available: run it directly (no council needed).

Step 1: Ground Truth Search (PARALLEL)

Call ALL available search tools in parallel:

• mcp__tachibot-mcp__grok_search (preferred - real-time)
• mcp__tachibot-mcp__perplexity_ask (citations, academic)
• mcp__tachibot-mcp__openai_search (GPT grounding)
• mcp__tachibot-mcp__gemini_search (Google grounding)

Skip any that aren't available. If NO search tools available, proceed without search grounding and note it.

Step 2: Parallel Analysis (all available)

Call ALL available analysis tools in parallel:

Reasoning (pick all available):

• mcp__tachibot-mcp__grok_reason - first principles analysis
• mcp__tachibot-mcp__kimi_thinking - step-by-step, edge cases
• mcp__tachibot-mcp__deepseek_reason - frontier open-weight reasoning/math
• mcp__tachibot-mcp__glm_reason - agentic reasoning (SWE-Bench Pro leader)

Code (if query involves code):

• mcp__tachibot-mcp__qwen_coder - code review and bugs
• mcp__tachibot-mcp__qwen_algo - algorithm optimization

Step 3: First Judge (with Pre-Mortem)

Use the FIRST available from this priority list:

1. mcp__tachibot-mcp__openai_reason (GPQA 87.7%)
2. mcp__tachibot-mcp__gemini_analyze_text (1M context)
3. mcp__tachibot-mcp__grok_reason (strong reasoning)

Prompt: "Extract key insights from each expert, identify consensus and conflicts. Then apply PRE-MORTEM: assume the recommended approach FAILED — what were the top 3 causes? Are those causes addressed in the analysis? Provide preliminary verdict with failure risks."

Step 4: Final Synthesis

Use a DIFFERENT model from Step 3. Priority:

1. mcp__tachibot-mcp__gemini_analyze_text (if not used in step 3)
2. mcp__tachibot-mcp__openai_reason (if not used in step 3)
3. Skip if only 1 model available (step 3 IS the final answer)

Prompt: Review first judgment, synthesize best elements from ALL analyses, resolve conflicts, provide final answer with confidence level.

Step 5: Present Results

Format:

• Answer: 1-2 sentence summary
• Models Used: [list which were available]
• Expert Insights: Key finding per model
• Consensus: Agreed points
• Conflicts Resolved: How settled
• Recommendation: Next steps
• Confidence: High/Medium/Low

Examples

• /judge how to implement rate limiting
• /judge analyze this code: [paste]
• /judge microservices vs monolith for 10M users