Documentation Verifier

Philosophy

Generating documentation is a solved problem. Preventing documentation decay is not. This skill validates that generated docs remain accurate, linked, and fresh over time.

Verification Checks

1. Markdown Lint

Check heading hierarchy and structural consistency across all .md files. Verify: heading hierarchy (no H1 > H3 skips — critical for RAG/GEO), single H1 per doc, consistent formatting (trailing whitespace, list markers, blank lines around headings), no bare URLs.

2. Link Validation

Check all internal and external links. Internal: verify target file exists, anchor links match headings, check case-sensitivity (Linux vs macOS). External: check HTTP status, timeout 10s, skip authenticated URLs, flag 404s/5xx. Enhanced detection patterns (case-sensitivity, fragments, redirect chains, nested relative links) in SKILL-reference.md.

3. llms.txt Sync Check

Verify llms.txt references match actual files on disk. Check for: missing files, orphaned docs not listed in llms.txt, description drift (first paragraph check), stale llms-full.txt.

4. Image Validation

Check all image references in Markdown. Verify: file exists on disk, alt text present (flag empty ![]()), absolute URLs for registry-published packages (see SKILL-reference.md for platform URL patterns), file size under 1MB.

5. Freshness Check

Flag stale docs using git log last-modification dates. See SKILL-reference.md for per-file staleness thresholds. Compare against latest commit date, not calendar date — dormant projects with no commits shouldn't trigger warnings.

6. Feature Coverage Sync

Compare README features against actual code using feature-benefits skill extraction. Flag undocumented features (code evidence exists, not in README) and over-documented features (claimed but no code evidence).

7. Badge URL Validation

Verify shields.io badge URLs return valid responses (HTTP 200, not error SVGs). Flag broken badges and outdated URL formats.

Quality Score

After running all verification checks, calculate a numeric quality score. The score gives users a single number to track and improve — modelled on the grading approach used in documentation quality tooling across the ecosystem.

See SKILL-reference.md for the 6-dimension scoring rubric (Completeness, Structure, Freshness, Link Health, Evidence, GEO & Citation Readiness) and grade bands (A through F).

Score Calculation

score = 100
for each check result:
  apply deductions from SKILL-reference.md scoring dimensions
score = max(0, score)
grade = lookup(score)  // A: 90–100, B: 80–89, C: 70–79, D: 60–69, F: <60

Report: show per-dimension breakdown and always include an actionable "To reach next grade" suggestion with the 1-2 highest-impact fixes. See SKILL-reference.md for the full report format example.

Supports ci argument for pipeline use (/docs-verify ci --min-score 70) and --min-score N threshold. See SKILL-reference.md for CI score export snippets (GitHub Actions, GitLab CI).

8. Guide Frontmatter Validation

Verify docs/ files have valid YAML frontmatter per the user-guides skill standard. Required fields: title, description, type (one of tutorial, how-to, reference, explanation). Also check: title matches H1, related: paths exist on disk. Scoring: -2 per guide missing required frontmatter (Structure dimension).

9. Token Audit

Estimate token cost for all skill files in .claude/skills/ using wc -w × 1.3. Flag skills over 3,000 tokens (reference) or 5,000 tokens (combined). Full audit script and thresholds in SKILL-reference.md.

10. Security Scan

Scan docs for content that should never appear in public repos. Classify credential-pattern matches as: placeholder (acceptable), env var reference (acceptable), or real credential (block immediately). Also check for: leaked internal paths (/Users/, /home/, C:\Users\), internal hostnames/IPs (192.168., 10.0.), and non-existent package names (dependency confusion vectors).

11. AI Context Health (Lightweight)

Basic presence and staleness check for AI context files (CLAUDE.md, AGENTS.md, .cursorrules, etc.). Scoring: -2 per context file older than 90 days, -1 per missing file in the project's tool ecosystem. For full signal-gate scoring and line budget analysis, install ContextDocs.

CI Integration

When run with ci argument, output machine-readable ERROR:/WARN: lines with file:line format and exit code 1 on errors. Supports --min-score N threshold. Full CI output format and GitHub Actions workflow template in SKILL-reference.md.

Anti-Patterns

• Don't ignore warnings — a broken link today becomes a confused user tomorrow
• Don't run external link checks on every commit — run them on PRs and weekly schedules to avoid rate limiting
• Don't fix docs in a separate PR from code changes — docs updates should accompany the code that changes behaviour
• Don't suppress freshness warnings without reviewing — stale docs erode trust faster than missing docs