/moe-analyze - Multi-Agent Research with Certainty Quantification

Execute comprehensive multi-agent research using coordinated analyst subagents with explicit certainty scoring, evidence requirements, and logical inference documentation.

Usage

# Basic analysis
/moe-analyze "Analyze current DMS functional requirements for completeness"

# With specific focus
/moe-analyze --focus compliance "Review HIPAA and SOC2 alignment"

# With web search enabled
/moe-analyze --with-search "Research enterprise DMS market standards 2024-2025"

# Include judgement phase
/moe-analyze --with-judgement "Evaluate security implementation against OWASP Top 10"

# Specify certainty threshold
/moe-analyze --min-certainty 80 "Validate database schema design decisions"

System Prompt

System Prompt

⚠️ EXECUTION DIRECTIVE: When the user invokes this command, you MUST:

1. IMMEDIATELY execute - no questions, no explanations first
2. ALWAYS show full output from script/tool execution
3. ALWAYS provide summary after execution completes

DO NOT:

• Say "I don't need to take action" - you ALWAYS execute when invoked
• Ask for confirmation unless requires_confirmation: true in frontmatter
• Skip execution even if it seems redundant - run it anyway

The user invoking the command IS the confirmation.

---

You are executing the MoE Analysis Framework, coordinating multiple analyst agents to research a topic with explicit certainty scoring and evidence requirements.

Workflow

Phase 1: Analyst Dispatch (Parallel)

Dispatch 4-6 analyst agents simultaneously:

# Execute these Task calls in parallel

Task(
    subagent_type="web-search-researcher",
    description="Primary research gathering",
    prompt="""Research the topic: {user_topic}

    MANDATORY: For EACH finding, provide:
    1. Certainty Score (0-100%)
    2. Source URLs with reliability rating
    3. Source recency (publication date)
    4. Direct quotes or evidence
    5. Gaps in available information

    If no reliable source exists, explicitly state:
    "NO SUPPORTING DATA AVAILABLE"
    """
)

Task(
    subagent_type="thoughts-analyzer",
    description="Internal knowledge analysis",
    prompt="""Analyze existing internal documentation for: {user_topic}

    Search docs/, architecture files, and prior research.
    Rate certainty based on documentation quality and recency.
    """
)

Task(
    subagent_type="codebase-analyzer",
    description="Implementation verification",
    prompt="""Verify current implementation state for: {user_topic}

    Check actual code against documented requirements.
    Identify discrepancies with evidence.
    """
)

Task(
    subagent_type="qa-reviewer",
    description="Quality validation",
    prompt="""Review analysis quality for: {user_topic}

    Check for logical consistency.
    Identify unsupported claims.
    Flag overconfident assertions.
    """
)

Phase 2: Evidence Synthesis

Aggregate and cross-validate findings:

1. Collect all agent reports
2. Cross-reference overlapping claims
3. Identify contradictions requiring resolution
4. Calculate composite certainty scores

Phase 3: Certainty Scoring

For each finding, calculate:

certainty_score = (
    evidence_support * 0.40 +      # How well supported by sources
    source_reliability * 0.25 +    # Quality of sources
    internal_consistency * 0.20 +  # Agreement across agents
    recency * 0.15                 # How current the information
)

Certainty Levels:

Score	Level	Meaning
85-100%	HIGH	Strong evidence, reliable sources, consensus
60-84%	MEDIUM	Good evidence, some gaps or disagreement
30-59%	LOW	Limited evidence, significant uncertainty
0-29%	INFERRED	No direct evidence, logical inference only

Phase 4: Gap Documentation

MANDATORY: Explicitly document:

• What information was NOT found
• What sources were unavailable
• What would increase certainty
• Known limitations of analysis

Phase 5: Logical Inference (when data unavailable)

When evidence is insufficient, provide:

## Inferred Conclusion: [Statement]

**Inference Type:** [Deduction/Induction/Abduction]

**Reasoning Chain:**
1. Premise: [Statement] (Certainty: X%)
   Evidence: [Source or "Assumed"]
2. Premise: [Statement] (Certainty: X%)
   Evidence: [Source or "Assumed"]
3. Therefore: [Conclusion]

**Decision Tree:**

IF [condition A] is true (evidence: X) AND [condition B] is true (evidence: Y) THEN [conclusion C] follows WITH confidence: Z% ELSE [alternative conclusion]

**Inference Confidence:** [0-100%]
**Key Assumptions:** [List assumptions made]
**Falsification Criteria:** [What would disprove this]

Output Format

# MoE Analysis Report: [Topic]

**Generated:** [Timestamp]
**Analysts:** [List of agents used]
**Overall Certainty:** [X%] ([LEVEL])

---

## Executive Summary

[2-3 sentence summary with key certainty caveats]

---

## Findings

### Finding 1: [Title]

**Certainty:** [X%] ([LEVEL])

**Evidence:**
1. **Source:** [URL]
   - Reliability: [Peer-reviewed/Government/Industry/Blog/Unknown]
   - Date: [Publication date]
   - Quote: "[Relevant excerpt]"

2. **Source:** [URL]
   - ...

**Cross-Validation:** [How agents agreed/disagreed]

**Gaps:** [What's missing]

---

### Finding 2: [Title]
...

---

## Information Gaps

| Gap | Impact | What Would Help |
|-----|--------|-----------------|
| [Missing info] | [HIGH/MEDIUM/LOW] | [Suggested research] |

---

## Inferred Conclusions (Low Evidence)

### Inference 1: [Statement]

**Certainty:** [X%] (INFERRED)
**Reasoning Chain:** [Step-by-step logic]
**Decision Tree:** [Visual representation]
**Key Assumptions:** [List]

---

## Confidence Summary

| Finding | Certainty | Evidence Quality | Sources |
|---------|-----------|------------------|---------|
| [F1] | X% | [Rating] | [Count] |
| [F2] | X% | [Rating] | [Count] |

**Aggregate Certainty:** [X%]
**Lowest Confidence Area:** [Topic]
**Recommended Follow-up:** [Next steps]

Options

Option	Description
--focus [area]	Narrow research to specific domain
--with-search	Enable web search for external sources
--with-judgement	Add MoE judge evaluation phase
--min-certainty [N]	Require minimum certainty threshold
--output [path]	Write report to specific file
--agents [N]	Number of analyst agents (default: 4)

Action Policy

<default_behavior> This command EXECUTES research and PRODUCES deliverables:

• Dispatches analyst agents in parallel
• Gathers and synthesizes evidence
• Calculates certainty scores
• Documents gaps and limitations
• Generates comprehensive report

User receives actionable analysis with clear confidence levels. </default_behavior>

After analysis, verify: - All findings have certainty scores - All claims have evidence or are marked INFERRED - Gaps are explicitly documented - Logical inferences show reasoning chains - No overconfident assertions without evidence

---

Related Commands:

• /moe-judge - Evaluate and grade analysis quality
• /multi-agent-research - Standard research workflow
• /smart-research - Single-agent research

Related Skills:

• uncertainty-quantification - Certainty scoring patterns
• evaluation-framework - Quality assessment rubrics

Success Output

When moe-analyze completes:

✅ COMMAND COMPLETE: /moe-analyze
Topic: <analysis-topic>
Analysts: N dispatched
Certainty: N% (LEVEL)
Findings: N with evidence
Gaps: N documented

Completion Checklist

Before marking complete:

• Analysts dispatched
• Evidence gathered
• Certainty scored
• Gaps documented
• Report generated

Failure Indicators

This command has FAILED if:

• ❌ No topic provided
• ❌ No analysts responded
• ❌ No certainty scores
• ❌ Gaps not documented

When NOT to Use

Do NOT use when:

• Simple factual query
• Single source sufficient
• Time-critical (use /smart-research)

Anti-Patterns (Avoid)

Anti-Pattern	Problem	Solution
Skip certainty	Overconfidence	Always score certainty
Ignore gaps	False completeness	Document all gaps
Too few agents	Limited perspective	Use 4+ analysts

Principles

This command embodies:

• #9 Based on Facts - Evidence with certainty
• #6 Clear, Understandable - Clear confidence levels
• #3 Complete Execution - Full analysis workflow

Full Standard: CODITECT-STANDARD-AUTOMATION.md

Version: 1.0.0 Last Updated: 2025-12-19