Documentation Intelligence System

Overview

This skill enables systematic, parallelized analysis of documentation using multiple specialized agents. It maximizes throughput by running independent sections in parallel while respecting dependencies, then consolidates findings into actionable SOPs.

System Architecture

┌─────────────────────────────────────────────────────────────────────┐
│                      META-ORCHESTRATOR                               │
│  (Coordinates all phases, manages dependencies, tracks progress)     │
└─────────────────────────────────────────────────────────────────────┘
                                  │
          ┌───────────────────────┼───────────────────────┐
          ▼                       ▼                       ▼
┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│   DOC-OVERVIEW  │    │ PROMPT-ENGINEER │    │   CONSOLIDATOR  │
│  (Structure     │    │ (Optimize all   │    │  (Synthesize    │
│   discovery)    │    │  agent prompts) │    │   findings)     │
└─────────────────┘    └─────────────────┘    └─────────────────┘
          │                                            │
          ▼                                            ▼
┌─────────────────────────────────────────┐  ┌─────────────────┐
│         SECTION-ANALYZERS               │  │  SOP-GENERATOR  │
│  (Parallel execution per section)       │  │  (Create SOPs)  │
│  ┌────┐ ┌────┐ ┌────┐ ┌────┐ ┌────┐   │  └─────────────────┘
│  │ S1 │ │ S2 │ │ S3 │ │ S4 │ │... │   │
│  └────┘ └────┘ └────┘ └────┘ └────┘   │
└─────────────────────────────────────────┘

Workflow Phases

Phase 0: Structure Discovery

Agent: doc-overview Action: Scan documentation structure Output:

• Complete section/subsection map
• Dependency graph
• Parallelization tiers
• Agent assignments

Phase 1-N: Parallel Section Analysis

Agent: section-analyzer (multiple instances) Action: Deep-dive each section based on tier Parallelization: Launch all agents in current tier simultaneously Output: Structured analysis per section

Phase P: Prompt Optimization

Agent: prompt-engineer Action: Optimize prompts based on section findings Output: Refined prompts for future runs

Phase C: Consolidation

Agent: consolidator Action: Synthesize all section analyses Output:

• Unified knowledge model
• Cross-cutting patterns
• Integration synergies
• Knowledge gaps

Phase S: SOP Generation

Agent: sop-generator Action: Create procedures from consolidated knowledge Output: Standard Operating Procedures

Parallelization Strategy

Maximum Throughput Mode (20x subscription)

Launch all Tier 0 sections simultaneously (up to 20 agents)
As each completes, launch dependent Tier 1 sections
Continue until all tiers complete
Run consolidation
Generate SOPs in parallel for independent workflows

Dependency Rules

independent → Can run immediately
soft_dependency → Can run, will be enhanced by dependency
hard_dependency → Must wait for dependency to complete

Cross-Agent Communication

Shared Context Protocol

{
  "terminology": {
    "term": "definition agreed by all agents"
  },
  "cross_references": [
    {"from": "section_a", "to": "section_b", "type": "extends"}
  ],
  "questions": [
    {"from": "agent_x", "question": "...", "answered_by": "agent_y"}
  ]
}

Message Types

• TERM_DEFINITION: New terminology established
• CROSS_REF: Cross-reference discovered
• QUESTION: Question for other agents
• ANSWER: Answer to previous question
• CONFLICT: Conflicting information found
• SYNERGY: Integration opportunity identified

Usage Examples

Full Documentation Analysis

"Use doc-intelligence to analyze the Claude Code documentation completely,
identify all sections and dependencies, analyze each in parallel,
consolidate findings, and generate SOPs for common workflows."

Specific Section Deep-Dive

"Analyze the 'Build with Claude Code' section using doc-intelligence,
with special focus on Plugins and MCP integration patterns."

SOP-Focused Analysis

"Use doc-intelligence to create SOPs for setting up a new Claude Code
project with MCP servers and custom hooks."

Progress Monitoring

Status Dashboard Format

╔════════════════════════════════════════════════════════════╗
║ Documentation Intelligence Operation                        ║
╠════════════════════════════════════════════════════════════╣
║ Phase: 2 of 5 (Parallel Analysis)                          ║
║ Progress: ████████░░░░░░░░░░░░ 40%                        ║
║ Active Agents: 12/20                                        ║
║ Sections Complete: 8/20                                     ║
║ Est. Remaining: 15 minutes                                  ║
╠════════════════════════════════════════════════════════════╣
║ Current Tier: 1 (8 sections parallel)                      ║
║ Waiting: Tier 2 (4 sections, blocked on dependencies)      ║
╚════════════════════════════════════════════════════════════╝

Output Artifacts

After completion, you will have:

1. documentation-topology.md - Complete structure map
2. section-analyses/ - Individual section reports
3. optimized-prompts.md - Refined prompts for each section
4. consolidated-intelligence.md - Synthesized knowledge
5. sops/ - Standard Operating Procedures
6. CLAUDE.md updates - Integrated learnings

Continuous Improvement

Learning Loop

Execute → Capture patterns → Update prompts → Execute better
           ↑                                      │
           └──────────────────────────────────────┘

Revision Triggers

• New documentation version released
• Execution feedback received
• Knowledge gaps identified
• Better patterns discovered

Integration with Memory Systems

claude-mem Integration

• Store successful analysis patterns
• Remember cross-section learnings
• Cache optimized prompts
• Track improvement metrics

CLAUDE.md Updates

• Add discovered best practices
• Update agent configurations
• Record successful workflows
• Note areas for improvement