Documentation Health Audit
Cortex includes a five-phase documentation audit pipeline that assesses docs across structure, accuracy, completeness, quality, and information architecture. Run the full pipeline before releases, or use individual phases for targeted checks.
The Pipeline
Each phase builds on the previous. Structural issues are found before accuracy is checked, accuracy before completeness, and so on.
Phase 1: Structure → doc-maintenance
Phase 2: Accuracy → doc-claim-validator
Phase 3: Completeness → doc-completeness-audit
Phase 4: Quality → doc-quality-review
Phase 5: Architecture → doc-architecture-review
What Each Phase Checks
| Phase | Skill | Question it answers | Output |
|---|---|---|---|
| 1. Structure | doc-maintenance |
Are links working? Are there orphaned files? Is the folder structure clean? | Broken link list, orphan report, staleness scores |
| 2. Accuracy | doc-claim-validator |
Do the docs match the code? Are file paths, commands, and behavioral claims true? | Failed claims by severity (P0–P4) |
| 3. Completeness | doc-completeness-audit |
Is everything documented? What topics are missing for each audience? | Gap report with audience-weighted priorities |
| 4. Quality | doc-quality-review |
Is the writing clear, consistent, and appropriate for its audience? | Scored review (readability, consistency, audience fit, structure, actionability) |
| 5. Architecture | doc-architecture-review |
Can users find what they need? Is the hierarchy logical? | IA assessment with 7 heuristic scores |
Running the Full Pipeline
Invoke the orchestrator skill to run all five phases in sequence:
/ctx:doc-health-audit
The orchestrator handles phase ordering and gating — if Phase 1 finds critical structural issues, it warns you before proceeding to later phases that depend on a healthy foundation.
Output
The orchestrator produces a combined dashboard:
| Phase | Status | Key Metric | Top Finding |
|---|---|---|---|
| 1. Structure | PASS/WARN/FAIL | N broken links, N orphans | … |
| 2. Accuracy | PASS/WARN/FAIL | N P0 claims failed | … |
| 3. Completeness | PASS/WARN/FAIL | N% coverage, N P0 gaps | … |
| 4. Quality | PASS/WARN/FAIL | Grade A–F | … |
| 5. Architecture | PASS/WARN/FAIL | Grade A–F | … |
Plus a prioritized list of the top 5 actions across all phases.
Running Individual Phases
Each phase is a standalone skill you can invoke directly:
/ctx:doc-maintenance
/ctx:doc-claim-validator
/ctx:doc-completeness-audit
/ctx:doc-quality-review
/ctx:doc-architecture-review
Common Partial Runs
| Scenario | Phases |
|---|---|
| “Are the docs accurate after this refactor?” | 1 + 2 |
| “What docs are missing?” | 1 + 3 |
| “Is the writing good enough to publish?” | 4 only |
| “Can users find things after the reorg?” | 1 + 5 |
| “Quick pre-release check” | 1 + 2 + 3 |
Phase Details
Phase 1: Structure (doc-maintenance)
Scans for broken links, orphaned pages, empty files, stale timestamps, and folder structure compliance. This is the foundation — broken structure makes later phases unreliable.
Phase gate: If >10 broken internal links or critical structural issues are found, the orchestrator warns before continuing.
Phase 2: Accuracy (doc-claim-validator)
Extracts testable claims from markdown — file paths, commands, function references, config keys, behavioral assertions — and verifies them against the actual codebase.
Uses a two-phase approach:
- Deterministic verification for structural claims (file exists? command exists? symbol found?)
- AI-assisted verification for semantic claims (behavioral assertions, dependency claims, code examples)
Includes git staleness scoring: counts commits to referenced files after the doc was last edited to surface likely-stale content.
Phase gate: P0 findings (user-facing docs that would break if followed) trigger a warning.
Phase 3: Completeness (doc-completeness-audit)
Builds an inventory of what should be documented from four sources:
- Public code surface — CLI flags, API endpoints, config keys, exported functions
- User-facing features — TUI screens, workflows, integrations
- Operational surface — installation, upgrade paths, troubleshooting
- Cross-reference promises — “see also” links to pages that don’t exist
Maps each inventory item to existing docs and classifies as Documented, Shallow, Missing, or Misplaced. Gaps are prioritized by audience impact — a missing install guide (P0 for new users) outranks a missing advanced config reference (P2 for contributors).
Phase 4: Quality (doc-quality-review)
Scores each document across five dimensions:
| Dimension | What it measures |
|---|---|
| Readability | Sentence length, passive voice, jargon, paragraph density |
| Consistency | Terminology, formatting, heading style, tone |
| Audience Fit | Prerequisites, explanation depth, context gaps |
| Structure | Heading hierarchy, scannability, front-loading, section length |
| Actionability | Copy-pasteable examples, complete steps, expected output |
Dimensions are weighted by doc type — actionability matters more for tutorials than explanations, consistency matters more for reference pages.
Produces a letter grade (A–F) and identifies systemic issues that affect multiple docs.
Phase 5: Architecture (doc-architecture-review)
Evaluates information architecture against seven heuristics:
| Heuristic | What it assesses |
|---|---|
| Findability | Can readers locate content without knowing where it lives? |
| Hierarchy Coherence | Does the nesting make sense? Is it predictable? |
| Progressive Disclosure | Is content layered from simple to complex? |
| Cross-Linking | Do links create useful connections between pages? |
| Pattern Consistency | Do similar pages follow similar structures? |
| Separation of Concerns | Are reference, tutorial, guide, and explanation content distinct? |
| Maintenance Burden | Can the structure accommodate growth without reorganization? |
Includes navigation path analysis (tracing user journeys through the docs) and mental model comparison (does the structure match how users think about the product?).
When to Run
| Trigger | Recommended scope |
|---|---|
| Before a release | Full pipeline on all docs |
| After a major refactor | Phase 2 (accuracy) on affected docs |
| After restructuring docs | Phase 1 (structure) + Phase 5 (architecture) |
| New docs added | Phase 3 (completeness) + Phase 4 (quality) |
| Users report confusion | Phase 4 (quality) + Phase 5 (architecture) |
| Monthly health check | Full pipeline |
All Documentation Skills
Production
| Skill | Command | Purpose |
|---|---|---|
documentation-production |
/ctx:documentation-production |
Generate or organize project documentation |
reference-documentation |
/ctx:reference-documentation |
Produce reference-oriented documentation |
tutorial-design |
/ctx:tutorial-design |
Design hands-on tutorials with exercises |
mermaid-diagramming |
/ctx:mermaid-diagramming |
Create Mermaid diagrams for docs |
Review & Audit
| Skill | Command | Purpose |
|---|---|---|
doc-health-audit |
/ctx:doc-health-audit |
Full pipeline: runs all five audit phases in order |
doc-maintenance |
/ctx:doc-maintenance |
Structural health (broken links, orphans, staleness) |
doc-claim-validator |
/ctx:doc-claim-validator |
Semantic accuracy (verify claims match code) |
doc-completeness-audit |
/ctx:doc-completeness-audit |
Topic coverage (inventory gaps by audience) |
doc-quality-review |
/ctx:doc-quality-review |
Prose quality (readability, consistency, audience fit) |
doc-architecture-review |
/ctx:doc-architecture-review |
Information architecture (findability, navigation, hierarchy) |
Related
- Documentation Commands —
cortex docsCLI for browsing bundled docs - Skills Guide — how skills work and how to discover them
- Skill Showcase — deep dives into the most sophisticated skills