Document Hub: Initialize

Bootstrap a new project with a complete documentation hub structure.

Helper Scripts Available:

• scripts/validate_hub.py - Validates documentation structure
• scripts/extract_glossary.py - Extracts domain-specific terms from code
• scripts/detect_drift.py - Detects existing modules and technologies

Always run scripts with --help or check scripts/README.md first to understand their usage and output format.

What This Skill Does

Creates the Documentation Hub file structure in a project's cline-docs/ directory:

project-root/
└── cline-docs/
    ├── systemArchitecture.md      # High-level architecture diagrams
    ├── keyPairResponsibility.md   # Module responsibilities
    ├── glossary.md                # Domain-specific terms
    └── techStack.md               # Technologies used

Decision Tree: When to Use This Skill

User wants to set up documentation → Does cline-docs/ exist?
    ├─ Yes → Run validation first
    │         ├─ Valid → Use /document-hub read instead
    │         └─ Invalid → Ask user: overwrite or skip?
    │
    └─ No → Initialize new hub:
        1. Create directory structure
        2. Detect technologies (detect_drift.py)
        3. Extract glossary terms (extract_glossary.py)
        4. Generate initial content
        5. Validate result (validate_hub.py)

Initialization Workflow

Step 1: Check Existing State

First, check if a documentation hub already exists:

# Check if cline-docs exists
ls project-root/cline-docs/

# If exists, validate it first
python scripts/validate_hub.py /path/to/project

If validation returns "valid": false with missing files, proceed with initialization.

Step 2: Gather Project Information

Use helper scripts to analyze the project:

Detect existing technologies and modules:

python scripts/detect_drift.py /path/to/project

This returns JSON with:

• Actual modules found in src/
• Technologies detected from package.json/requirements.txt
• Configuration files present

Extract domain-specific terms:

python scripts/extract_glossary.py /path/to/project

This returns ranked terms with context from code comments.

Step 3: Create Documentation Files

Create the four core files with initial content based on the gathered information.

Templates: starter templates for all four files live in references/templates/ (systemArchitecture.md, keyPairResponsibility.md, glossary.md, techStack.md). Read the relevant template file when creating each doc file.

Step 4: Populate with Detected Information

Use the JSON output from helper scripts to populate initial content:

1. From detect_drift.py output:

   • Add detected technologies to techStack.md
   • Add detected modules to keyPairResponsibility.md

2. From extract_glossary.py output:

   • Add top 20-30 terms to glossary.md alphabetically
   • Include contexts as definitions

Step 5: Prompt User for Additional Context

After creating initial files, ask the user:

Documentation hub initialized with detected information.

Please provide additional context:
1. What is the primary purpose of this project?
2. Are there any key architectural decisions to document?
3. Any specific modules or workflows that need detailed explanation?

I can update the documentation with your input.

Step 6: Validate Result

After initialization, always validate:

python scripts/validate_hub.py /path/to/project

Check that:

• "valid": true
• All required files exist
• No Mermaid syntax errors

Example: Complete Initialization

A full end-to-end Python example of the 6-step workflow is in references/example-initialization.md — read it if you want a concrete script-driven walkthrough.

Best Practices

• Run detection scripts first - Don't guess what's in the project
• Use templates - Ensure consistent structure across projects
• Validate after creation - Always run validate_hub.py at the end
• Prompt for context - Auto-detected info needs human input for completeness
• Keep diagrams simple - Initial architecture diagrams should be high-level

Common Pitfalls

❌ Don't create documentation without analyzing the codebase first ❌ Don't overwrite existing valid documentation without user confirmation ❌ Don't create empty files - populate with at least template structure

✅ Do use helper scripts to gather information ✅ Do validate before and after ✅ Do ask user for additional context

Bootstrapping both systems at once (formerly /documentation-start)

To initialize BOTH Brain systems in one pass: run this skill (skip if cline-docs/ already has all 4 files), then run /memory-bank-initialize (skip if memory-bank/ already has its 6 files). Re-run either to force re-initialization if a system is present but incomplete.

What Comes Next

After initialization:

1. User reviews generated documentation
2. User provides additional context via prompts
3. Documentation is refined
4. Future updates use /document-hub update skill

Helper Script Reference

Usage and JSON output shapes for validate_hub.py, detect_drift.py, and extract_glossary.py are in references/script-reference.md — read it before invoking a script; scripts/README.md has complete documentation.