Agent Factory Intelligence Engine

Master Skill for Intelligent Agent Generation

Complete reference for agent-factory agent to automatically generate production-ready
Claude Code sub-agents through requirement analysis, research, template generation,
and validation.

Version: 1.0.0 Status: Production Ready Components: 6 core systems + advanced features

🎯 Quick Start: The 6 Core Components

🚀 Agent Generation Workflow

User Requirement
    ↓
[Stage 1-2] Intelligence Engine
  ├─ Parse requirement
  ├─ Detect domain (primary + secondary)
  ├─ Score complexity (1-10)
  ├─ Select model (Sonnet/Haiku/Inherit)
  └─ Calculate tools & skills
    ↓
[Stage 3] Research Engine (Context7 MCP)
  ├─ Resolve libraries
  ├─ Fetch documentation
  ├─ Extract best practices
  └─ Synthesize evidence
    ↓
[Stage 4-5] Template System
  ├─ Select template tier (1-3)
  ├─ Generate agent markdown
  └─ Apply variable substitution
    ↓
[Stage 6] Validation Framework
  ├─ Gate 1: YAML syntax
  ├─ Gate 2: Structure completeness
  ├─ Gate 3: Content quality
  └─ Gate 4: TRUST 5 + Claude Code v4.0
    ↓
Production-Ready Agent ✅

📚 Understanding Each Component

Intelligence Engine

Responsible for: Requirement analysis and decision making

Key Algorithms:

• Requirement Analysis: Extract domain, capabilities, complexity
• Domain Detection: Primary + secondary domains with confidence
• Complexity Scoring: 1-10 scale for model selection
• Model Selection: Sonnet/Haiku/Inherit decision tree
• Tool Calculator: Minimum necessary permissions
• Skill Recommender: Match to 128+ MoAI-ADK skills

See: [reference/intelligence-engine.md](reference/intelligence-engine.md)

Research Engine

Responsible for: Fetching official documentation and best practices

Key Workflows:

• Library Resolution: Convert domain → Context7 library IDs
• Documentation Fetch: Pull official docs from Context7 MCP
• Best Practice Extraction: Identify actionable patterns
• Pattern Identification: Architectural patterns for domain
• Quality Validation: Ensure research meets 70%+ threshold
• Evidence Synthesis: Consolidate into recommendations
• Fallback Strategy: WebFetch if Context7 unavailable

See: [reference/research-engine.md](reference/research-engine.md)

Template System

Responsible for: Generating agent markdown with correct structure

Key Components:

• Tier 1 (Simple): ~200 lines, Haiku, <5 min generation
• Tier 2 (Standard): 200-500 lines, Inherit/Sonnet, <15 min
• Tier 3 (Complex): 500+ lines, Sonnet, 20-30 min
• Variables: 15+ categories for dynamic substitution
• YAML Generation: Frontmatter with proper metadata

Template Files (in templates/ subfolder):

• simple-agent.template.md – Tier 1 template for simple agents
• standard-agent.template.md – Tier 2 template for standard agents
• complex-agent.template.md – Tier 3 template for complex agents
• VARIABLE_REFERENCE.md – Complete variable substitution guide

See: [reference/template-system.md](reference/template-system.md)

Validation Framework

Responsible for: Ensuring agent quality meets all standards

4 Quality Gates:

1. Gate 1: YAML syntax validation
2. Gate 2: Required sections verification
3. Gate 3: Content quality checks
4. Gate 4: TRUST 5 + Claude Code v4.0 compliance

Includes: 5 core test cases + 3 edge cases

See: [reference/validation-framework.md](reference/validation-framework.md)

Advanced Features

Responsible for: Enterprise capabilities and optimization

Features:

• Semantic Versioning: Track agent versions
• Multi-Domain Support: 2-3 domain agents
• Performance Analyzer: Automatic optimization suggestions
• Enterprise Compliance: SOC2, GDPR, HIPAA support
• Audit Logging: Activity tracking and monitoring

See: [reference/advanced-features.md](reference/advanced-features.md)

🔗 Integration Points

With agent-factory Agent

---
name: agent-factory
model: sonnet
---

## Required Skills
Skill("moai-alfred-agent-factory")  # This master skill

# In execution:
1. Load this skill
2. Use Intelligence Engine (Stage 1-2)
3. Use Research Engine (Stage 3)
4. Use Template System (Stage 4-5)
5. Use Validation Framework (Stage 6)

With @agent-cc-manager

After generation, delegate to @agent-cc-manager for:

• Claude Code v4.0 compliance verification
• .claude/settings.json validation
• MCP server configuration check
• Hook registration validation

With @agent-mcp-context7-integrator

Research Engine delegates to for:

• Library ID resolution
• Official documentation fetching
• Best practice identification
• Latest API version discovery

📊 Performance Expectations

🧠 How to Use This Skill

For Agent-Factory Agent

# 1. Load this skill
skill = Skill("moai-alfred-agent-factory")

# 2. Use Intelligence Engine for analysis
domain = skill.intelligence_engine.detect_domain(user_input)
complexity = skill.intelligence_engine.complexity_score(domain, capabilities)
model = skill.intelligence_engine.select_model(complexity)

# 3. Use Research Engine for documentation
research = skill.research_engine.research_domain(domain, frameworks)
practices = research.best_practices
patterns = research.patterns

# 4. Use Template System for generation
template = skill.template_system.select_template(complexity)
agent_markdown = skill.template_system.generate_agent(
    template=template,
    variables={...}
)

# 5. Use Validation Framework for quality
validation = skill.validation_framework.validate(agent_markdown)
if validation.passed:
    return agent_markdown
else:
    return validation.issues

For Reference Lookup

Need specific information? Use the reference files:

• "What model should I pick?" → See [intelligence-engine.md](reference/intelligence-engine.md#model-selection)
• "How do I get best practices?" → See [research-engine.md](reference/research-engine.md#best-practice-extraction)
• "What variables exist?" → See [template-system.md](reference/template-system.md#variables)
• "How are agents validated?" → See [validation-framework.md](reference/validation-framework.md#quality-gates)

📖 Reference Files Organization

moai-alfred-agent-factory/
├── SKILL.md                        (this file - navigation hub)
├── reference/                      (detailed documentation)
│   ├── intelligence-engine.md      (280 lines: algorithms)
│   ├── research-engine.md          (320 lines: workflows)
│   ├── template-system.md          (280 lines: templates)
│   ├── validation-framework.md     (220 lines: testing)
│   └── advanced-features.md        (200 lines: enterprise)
├── templates/                      (reusable agent templates)
│   ├── simple-agent.template.md    (Tier 1: <200 lines, Haiku)
│   ├── standard-agent.template.md  (Tier 2: 200-500 lines)
│   ├── complex-agent.template.md   (Tier 3: 500+ lines)
│   └── VARIABLE_REFERENCE.md       (15+ variable categories)
├── examples.md                     (719 lines: use cases)
└── reference.md                    (400 lines: quick lookup)

✨ Key Highlights

✅ Comprehensive: 6 core systems with complete documentation ✅ Modular: Each system independently referenceable ✅ Practical: Includes algorithms with code examples ✅ Tested: 5 core + 3 edge case test scenarios ✅ Enterprise: Versioning, compliance, optimization ✅ Official: Follows Claude Code v4.0 standards

🎓 Learning Path

New to agent-factory?

1. Read this overview
2. Review [examples.md](examples.md) for practical cases
3. Dive into specific reference files as needed

Integrating with your agent?

1. Load this skill with Skill("moai-alfred-agent-factory")
2. Reference specific algorithms from this overview
3. Consult detailed docs in reference/ folder

Need specific feature? Use the component table above and jump to the corresponding reference file.

📞 Quick Reference

Created: 2025-11-15 Version: 1.0.0 Status: Production Ready Total Content: 2,800+ lines across organized reference files