Whole Content Analyzer

Purpose

Run comprehensive analysis on Whole documentation sections before editing begins.

Integration with Agents

When to Invoke Agents

Use Task tool to invoke specialized analysis agents for deep analysis:

// For deep duplicate semantic analysis
Task(subagent_type: 'whole-translator',
     prompt: 'Analyze semantic similarity and cultural context for potential duplicates in CF[N]')

// For cross-reference graph analysis before editing
Task(subagent_type: 'whole-cross-reference',
     prompt: 'Build reference graph and analyze connectivity patterns for CF[N]')

// For comprehensive content structure validation
Task(subagent_type: 'whole-content-validator',
     prompt: 'Pre-validate content structure and identify gaps in CF[N]')

When NOT to Use Agents

• Simple concept counting → Use Grep directly
• Basic duplicate detection (exact matches) → Use Grep with concept names
• Format checking → Use shared utilities from .claude/skills/shared
• Quick inventory → Use Grep pattern matching

Analysis Types

1. Content Inventory

• Count concepts per section
• Identify incomplete 4-point descriptions
• Count cross-references

2. Duplicate Detection

• Exact matches across domains
• Similar concepts (>70% overlap)
• Classify: meaningful diversity vs redundancy

3. Structural Check

• Domain-function alignment
• Distribution across 5 functions
• Bilingual format compliance
• Cross-reference integrity

4. Gap Analysis

• Missing function categories
• Incomplete descriptions
• Weak integration points

Analysis Workflow

Phase 1: Quick Analysis (Scripts & Grep)

Run automated analysis first for basic metrics:

# Basic structure and format checks
grep -n "^#### \*\*[0-9]" Whole.md | wc -l  # Count concepts
grep -n "→ \*\*Liên kết:" Whole.md | wc -l   # Count cross-refs

Phase 2: Agent-Based Deep Analysis

For complex analysis requiring semantic understanding:

1. Invoke whole-content-validator - Get comprehensive structure validation
2. Invoke whole-cross-reference - Build reference graph and identify patterns
3. Invoke whole-translator (if needed) - Semantic duplicate detection

Phase 3: Report Generation

Synthesize findings from both automated and agent-based analysis.

Agent Integration Guide

whole-content-validator

When to use: Pre-validate structure and identify gaps before editing Command: Task(subagent_type='whole-content-validator', prompt='Pre-validate CF[N]') Expected output: Structure validation report with gap analysis

whole-cross-reference

When to use: Analyze reference patterns and identify weak/strong connectivity Command: Task(subagent_type='whole-cross-reference', prompt='Analyze reference patterns in CF[N]') Expected output: Reference graph, connectivity analysis, strategic recommendations

whole-translator

When to use: Detect semantic duplicates across bilingual concepts Command: Task(subagent_type='whole-translator', prompt='Analyze semantic similarity for duplicates in CF[N]') Expected output: Semantic similarity report, cultural context analysis

Output Format

# Analysis Report: CHỨC NĂNG [N]

## Summary
- Total Concepts: [N]
- Complete: [N] ([%])
- Cross-Refs: [N]
- Issues: [N]

## Automated Analysis Results

### Quick Metrics:
- Concept count: [N]
- Cross-references: [N]
- Format compliance: [%]

### Agent Analysis:
- ✅ **whole-content-validator**: [Summary]
- ✅ **whole-cross-reference**: [Summary]
- ⚠️ **whole-translator**: [Summary]

## Findings

### Strengths
[List]

### Issues
[List with severity: Critical/Warning/Info]

### Potential Duplicates
[Table: name, locations, similarity %, classification, agent recommendation]

### Connectivity Analysis
- High-connectivity concepts: [List]
- Isolated concepts: [List]
- Recommended new links: [List]

## Recommendations
[Prioritized actions based on both automated and agent findings]

---
**Analyzer**: whole-analyzer v2.1.0
**Agents Invoked**: [List]
**Date**: [timestamp]

Critical Rules

✅ MUST

• Run quick analysis first before invoking agents
• Use agents for deep semantic analysis, not simple checks
• Document agent findings in analysis report
• Use shared utilities from .claude/skills/shared
• Provide actionable recommendations

❌ NEVER

• Invoke agents for simple metrics (use Grep/scripts)
• Skip quick analysis phase
• Modify content (analysis only)
• Analyze without clear purpose