creo-seo-content — Content Quality Specialist

Deep content auditor for Next.js apps. Scores each page across 5 content dimensions, detects AI-generated patterns, maps search intent, scores opportunity, and outputs actionable rewrites with before/after examples.

Configuration load

1. Read .claude/project-config.md for project_id, reports_path, keyword strategy
2. Load project profile: .claude/skills/creo-seo/creo-seo-{project_id}.md (shared with creo-seo skill)
3. If called by creo-seo as Phase 2 delegate, accept inherited context (URL list, business type, project config)

Reference library

File	Purpose
skills/creo-seo-content/references/search-intent.md	4-class intent system
skills/creo-seo-content/references/opportunity-scoring.md	8-factor weighted scoring
skills/creo-seo-content/references/content-playbook.md	15 writing patterns
skills/creo-seo-content/references/rewrite-templates.md	Before/after examples
skills/creo-seo/references/content-quality-rubric.md	5-dim content score
skills/creo-seo/references/ai-pattern-detection.md	26 AI phrases + 24 vague words
skills/creo-seo/references/geo-citability.md	5-dim citability + 134-167w window
skills/creo-seo/references/internal-linking-recipe.md	Linking rules

Audit flow per page

1. Fetch content

• Local MDX/TSX: Read file
• Live URL: WebFetch and extract main content (skip nav/footer/sidebar)
• Normalize to plain text + structure map (headings, paragraphs, lists, tables)

2. Score 5 dimensions

Apply content-quality-rubric.md:

Content_Quality =
  (Humanity        × 0.30) +
  (Specificity     × 0.25) +
  (Structure       × 0.20) +
  (SEO_Compliance  × 0.15) +
  (Readability     × 0.10)

3. AI-pattern sweep

Apply ai-pattern-detection.md:

• Count every AI phrase occurrence (line-numbered)
• Count every vague quantifier
• Count specificity boosters (%, $, years, citations)
• Count conversational boosters (contractions, questions, asides)
• Compute per-500-word densities
• Compute humanity score

4. Structure analysis

• Heading hierarchy check (single H1, logical H2→H3)
• Paragraph length distribution
• Table / list presence for comparable content
• Question-format H2 coverage

5. Citability scoring

Apply geo-citability.md:

• Segment content into passages around each H2/H3
• Find passages in 134–167 word range
• Score each on 5 dimensions (answer quality, self-containment, readability, stats density, uniqueness)
• Aggregate to page citability score

6. Intent classification

Apply search-intent.md:

• Extract keyword signals from title, H1, URL
• Infer primary + secondary intent
• Flag intent/content mismatch

7. Opportunity scoring (if GSC data available)

Apply opportunity-scoring.md 8-factor formula. Tier: CRITICAL / HIGH / MEDIUM / LOW / SKIP. Project traffic gain if position improves.

8. Internal linking audit

Apply internal-linking-recipe.md:

• Count outbound internal links
• Verify placement priority (contextual > intro > conclusion > list)
• Check anchor diversity (exact / partial / branded / natural)
• Validate cluster membership (does cluster article link up to pillar?)
• Flag orphans

9. Generate rewrite priorities

For each problem detected, match to a template in rewrite-templates.md. Generate:

• Problem description
• BEFORE (current content with line numbers)
• AFTER (concrete rewrite)
• Projected score lift
• File + line location to apply

Rank by projected score impact.

10. Write report

Save to {reports_path}/seo-content-{slug}-{YYYYMMDD}-{HHMM}.md. Include:

• Composite score + dimension breakdown
• AI pattern flags (line-numbered)
• Intent classification (primary + secondary)
• Opportunity score + tier
• Internal linking gaps
• Top 3–5 rewrite priorities with BEFORE/AFTER
• Next steps

Output rules

• Every AI phrase flag includes line number
• Every rewrite priority includes BEFORE + AFTER
• File paths use project's actual content location (app/blog/[slug]/page.mdx, content/posts/*.mdx, etc.)
• Projected score lift is calculated, not hand-waved
• Publish gate output: "PASS" (≥ 80 + no Critical) or "NEEDS WORK" with specific fixes

Publish gate mode

When called with /creo seo-content gate <file>:

• Exit code 0 = pass (safe to publish)
• Exit code 1 = fail (score < 80 or Critical issues)
• Output: one-line verdict + top-3 blockers if fail

Useful as CI check on blog post PRs.

Integration with creo-seo

When spawned by creo-seo Phase 2:

• Receive URL list, business type, project profile
• Run steps 1–8 in parallel across URLs
• Return aggregated JSON with per-page scores + flags
• creo-seo merges into composite audit report

When run standalone:

• Accept single URL or glob pattern
• Run steps 1–10 serial
• Produce per-page markdown reports

Quality gates

• Every score reproducible from refs (no magic constants in prose)
• Every flag line-numbered
• Every rewrite concrete, not abstract
• Reports saved to disk, never console-only
• No recommendations without a corresponding reference-file citation