Document Hub: Analyze

Analyze documentation quality and detect drift from the actual codebase. This is a read-only diagnostic: report problems and recommendations, make no changes. Suggest /document-hub update if drift is detected.

Helper Scripts — canonical copies live in /home/artsmc/.claude/skills/document-hub-initialize/scripts/ (see its README.md for full docs):

• detect_drift.py <project> - module + technology drift (JSON: drift scores + specific gaps)
• extract_glossary.py <project> - ranked undocumented domain terms with contexts
• validate_hub.py <project> - structure validation (JSON: structural issues)

Analysis Workflow

Run all six phases, then present findings without making changes.

Phase 1: Validation Check

python /home/artsmc/.claude/skills/document-hub-initialize/scripts/validate_hub.py /path/to/project

If validation fails: report structural errors, recommend fixing before analyzing content, and exit early if the hub doesn't exist.

Phase 2: Module Drift

python /home/artsmc/.claude/skills/document-hub-initialize/scripts/detect_drift.py /path/to/project

Returns:

{
  "drift_score": 0.23,
  "module_drift": {
    "undocumented": ["analytics", "webhooks"],
    "documented_but_missing": ["legacy"]
  }
}

• undocumented → code exists (src/) but not in keyPairResponsibility.md
• documentedbutmissing → docs reference non-existent code

Phase 3: Technology Drift

Same detect_drift.py output:

{
  "technology_drift": {
    "undocumented": ["Redis", "BullMQ"],
    "documented_but_missing": ["MongoDB"]
  }
}

Check package.json/requirements.txt vs techStack.md; identify when tech was added (git log) and whether it is actually in use.

Phase 4: Glossary Gaps

python /home/artsmc/.claude/skills/document-hub-initialize/scripts/extract_glossary.py /path/to/project

Returns ranked terms from the codebase. Read the existing cline-docs/glossary.md, identify terms in code but not in the glossary, rank by the script's importance score, and recommend the top 10-20 additions.

Phase 5: Health Scoring

Health Score = 100 - (drift_score * 100)

Regardless of score, immediate action is needed for drift > 0.35, multiple undocumented core modules, or broken cross-references.

Phase 6: Recommendations

Prioritize by impact:

• HIGH: undocumented modules with many files; missing critical technologies (databases, frameworks); broken cross-references
• MEDIUM: documented-but-missing code; complex diagrams needing split; missing glossary terms for core concepts
• LOW: minor tech stack updates; formatting inconsistencies; additional glossary terms

Report Format

The full report template (section structure with sample output) and a complete Python example that runs all three scripts and assembles the report live in references/report-format.md — read it before presenting findings.

Use Cases

• Pre-update health check - run before /document-hub update to understand scope and prioritize changes
• Periodic audit (monthly/quarterly) - track health score trend, address high-priority items
• Onboarding validation - when taking over a project, gauge documentation completeness and identify knowledge gaps

Best Practices & Pitfalls

• ✅ Run before updating; use findings to guide updates; track health score trends over time
• ✅ Prioritize high-impact recommendations - don't try to fix everything at once
• ❌ Don't make changes during analysis - this is read-only
• ❌ Don't ignore high-priority items - they indicate real gaps
• ❌ Don't run too frequently - analysis is for planning, not continuous monitoring