small fix, don't worry about it

2026-06-05 04:53:01 +00:00 · 2026-04-10 12:06:24 +02:00
commit 5c500a72b0
243 changed files with 44176 additions and 0 deletions
@@ -0,0 +1,114 @@
+---
+name: seo
+description: Comprehensive SEO analysis for websites and businesses. Audit technical SEO, content quality (E-E-A-T), schema markup, sitemaps, Core Web Vitals with INP, image optimization, and AI search readiness for AI Overviews, ChatGPT, and Perplexity. Use when requests mention SEO, audit, schema, Core Web Vitals, sitemap, technical SEO, content quality, E-E-A-T, GEO, AI crawler access, llms.txt, programmatic SEO, competitor pages, or hreflang.
+---
+
+# SEO
+
+Perform comprehensive SEO analysis across SaaS, local services, e-commerce, publishers, agencies, and mixed business models. Run full audits or focused deep-dives, then return prioritized actions with clear severity.
+
+## Command Map
+
+| Command | Action |
+|---------|--------|
+| `/seo audit <url>` | Run full website audit with parallel workstreams |
+| `/seo page <url>` | Run deep single-page analysis |
+| `/seo sitemap <url or generate>` | Audit or generate XML sitemap strategy |
+| `/seo schema <url>` | Detect, validate, and generate schema |
+| `/seo images <url>` | Audit image SEO and media delivery |
+| `/seo technical <url>` | Audit crawlability/indexability/infra |
+| `/seo content <url>` | Evaluate E-E-A-T and content quality |
+| `/seo geo <url>` | Audit AI Overviews and GEO readiness |
+| `/seo plan <business-type>` | Produce strategic SEO roadmap |
+| `/seo programmatic [url|plan]` | Analyze programmatic SEO viability |
+| `/seo competitor-pages [url|generate]` | Build competitor comparison page plan |
+| `/seo hreflang [url]` | Audit hreflang and i18n SEO coverage |
+
+## Full Audit Workflow
+
+For `/seo audit <url>`, execute this order:
+1. Detect business type from homepage signals.
+2. Run six tracks in parallel when possible:
+   - Technical SEO
+   - Content quality
+   - Schema
+   - Sitemap
+   - Performance
+   - Visual/mobile UX
+3. Compute unified SEO Health Score.
+4. Return findings sorted by priority:
+   - Critical: fix now
+   - High: fix within 1 week
+   - Medium: fix within 1 month
+   - Low: backlog
+5. Provide a concrete action plan with expected impact.
+
+## Business Type Detection
+
+Infer primary business type using homepage and IA signals:
+- SaaS: `/pricing`, `/features`, `/integrations`, `/docs`, "free trial", "sign up"
+- Local service: address, phone, map embed, service areas, city modifiers
+- E-commerce: `/products`, `/collections`, cart flows, Product schema
+- Publisher: `/blog`, `/articles`, author pages, publish/update dates
+- Agency: `/case-studies`, `/portfolio`, client logos, industry pages
+
+If mixed signals exist, mark "hybrid" and tailor recommendations per page cluster.
+
+## Quality Gates
+
+Apply these non-negotiable rules:
+- Warn at 30+ location pages and require at least 60% unique content per page.
+- Hard stop at 50+ location pages unless user explicitly justifies scale and differentiation plan.
+- Never recommend HowTo schema as an SEO lever.
+- Recommend FAQ schema only for government and healthcare websites.
+- Use INP for responsiveness; do not use FID in recommendations.
+
+Load `references/quality-gates.md` before auditing large template sets.
+
+## Scoring Model
+
+Use weighted aggregate:
+
+| Category | Weight |
+|----------|--------|
+| Technical SEO | 25% |
+| Content Quality | 25% |
+| On-Page SEO | 20% |
+| Schema / Structured Data | 10% |
+| Performance (CWV) | 10% |
+| Images | 5% |
+| AI Search Readiness | 5% |
+
+Round final score to nearest integer and explain largest negative contributors.
+
+## Reference Loading Rules
+
+Do not load every reference at startup. Load only what the task needs:
+- `references/cwv-thresholds.md`: CWV threshold checks and measurement caveats
+- `references/schema-types.md`: supported/deprecated schema and eligibility notes
+- `references/eeat-framework.md`: E-E-A-T evaluation rubric
+- `references/quality-gates.md`: thin content and scaled-page gating rules
+
+## Sub-Skills and Delegation
+
+Use these capability labels in reports:
+1. `seo-audit`
+2. `seo-page`
+3. `seo-technical`
+4. `seo-content`
+5. `seo-schema`
+6. `seo-images`
+7. `seo-sitemap`
+8. `seo-geo`
+9. `seo-plan`
+10. `seo-programmatic`
+11. `seo-competitor-pages`
+12. `seo-hreflang`
+
+When splitting a full audit, delegate these tracks in parallel:
+- `seo-technical`
+- `seo-content`
+- `seo-schema`
+- `seo-sitemap`
+- `seo-performance`
+- `seo-visual`
@@ -0,0 +1,4 @@
+interface:
+  display_name: "SEO"
+  short_description: "Analyze websites for technical and content SEO issues"
+  default_prompt: "Run a full SEO audit and prioritized action plan for this website"
@@ -0,0 +1,50 @@
+# Core Web Vitals Thresholds (INP-era)
+
+Use field data first (CrUX/RUM), then lab data for diagnosis.
+
+## Primary Metrics
+
+| Metric | Good | Needs Improvement | Poor |
+|--------|------|-------------------|------|
+| LCP | <= 2.5s | > 2.5s and <= 4.0s | > 4.0s |
+| INP | <= 200ms | > 200ms and <= 500ms | > 500ms |
+| CLS | <= 0.10 | > 0.10 and <= 0.25 | > 0.25 |
+
+## Supporting Diagnostics
+
+Use these as diagnostic hints, not ranking stand-ins:
+
+| Metric | Suggested Target |
+|--------|------------------|
+| TTFB | <= 0.8s |
+| FCP | <= 1.8s |
+| Total Blocking Time (lab) | <= 200ms |
+| Main-thread long tasks | < 50ms each where possible |
+
+## Measurement Hierarchy
+
+1. Field data source priority:
+   - CrUX (origin + URL level if available)
+   - First-party RUM
+2. Lab data source priority:
+   - PageSpeed Insights lab profile
+   - Lighthouse (mobile + throttled) for reproducible debugging
+
+## Audit Instructions
+
+1. Compare mobile first, then desktop.
+2. Classify each page template by CWV risk:
+   - Critical: poor in any primary metric on key template
+   - High: needs improvement on key conversion template
+   - Medium: issue on secondary templates
+3. Map root causes:
+   - LCP: render-blocking CSS/JS, slow HTML, heavy hero image
+   - INP: long JS tasks, expensive handlers, third-party script contention
+   - CLS: unsized media, ad/embed shifts, late DOM injections
+4. Propose one quick win and one structural fix per failing metric.
+
+## Rules
+
+- Always use INP. Do not reference FID as a current optimization target.
+- Avoid over-claiming from synthetic tests alone.
+- If field data is unavailable, clearly label findings as provisional.
@@ -0,0 +1,73 @@
+# E-E-A-T Evaluation Framework
+
+Use this rubric to evaluate content quality for competitive queries.
+
+## Dimensions
+
+Score each dimension from 0 to 5:
+
+| Dimension | What to Check |
+|-----------|---------------|
+| Experience | First-hand evidence, practical examples, real usage context |
+| Expertise | Subject competence, technical accuracy, depth, specialist language used correctly |
+| Authoritativeness | Reputation signals, citations, recognized brand/person entities |
+| Trustworthiness | Accuracy, transparency, policy clarity, editorial rigor, safety disclosures |
+
+## Query Stakes Classification
+
+Classify intent before scoring:
+
+- YMYL-high: finance, health, legal, safety
+- Commercial-high: high-ticket buying decisions
+- Informational-competitive: dense SERP competition with strong entities
+- Informational-low: low-risk, basic educational intent
+
+Raise trust requirements for higher-stakes intents.
+
+## Evidence Checklist
+
+Review per template:
+
+1. Authorship clarity:
+   - named author
+   - credentials or role
+   - updated timestamp
+2. Source quality:
+   - primary sources cited
+   - external corroboration
+   - quote/context integrity
+3. Editorial quality:
+   - original analysis (not paraphrased commodity text)
+   - internal consistency across pages
+   - clear claims with evidence
+4. Trust assets:
+   - about page and editorial policy
+   - contact and business identity
+   - privacy and terms links
+5. User value:
+   - intent match
+   - completeness
+   - actionable depth
+
+## Risk Flags
+
+Flag as high risk when you find:
+- anonymous or credential-free expert claims on high-stakes pages
+- copied or near-duplicate boilerplate across key landing pages
+- medical/financial/legal recommendations without sourcing
+- exaggerated claims without verifiable proof
+
+## Scoring Output Format
+
+Return:
+1. Overall content quality score (0-100)
+2. Per-dimension scores mapped from rubric
+3. Top 3 trust risks
+4. Fastest high-impact remediation actions
+
+## Remediation Patterns
+
+- Add expert-reviewed blocks with named reviewers on high-stakes pages.
+- Replace generic intros with first-hand evidence and data-backed claims.
+- Add citation discipline: primary sources first, then secondary context.
+- Standardize author bylines, update dates, and revision ownership.
@@ -0,0 +1,62 @@
+# Content Quality Gates
+
+Use these gates before recommending scale, templates, or programmatic rollout.
+
+## Thin Content Thresholds
+
+Apply minimum effective content targets by page type:
+
+| Page Type | Minimum Body Content | Minimum Unique Content | Notes |
+|-----------|----------------------|------------------------|-------|
+| Homepage | 500 words | 70% | Differentiate value proposition and audience |
+| Service page | 700 words | 70% | Include process, proof, and FAQs |
+| Location page | 600 words | 60% | Require city-specific proof and context |
+| Product category | 350 words | 65% | Add buyer guidance, not filler |
+| Product detail | 250 words | 80% | Unique specs, differentiators, and use cases |
+| Blog/article | 900 words | 75% | Include original insights and references |
+| Competitor comparison | 1000 words | 80% | Require side-by-side evidence |
+
+If intent is fully satisfied with less copy, allow exception only with strong evidence (high engagement and conversion outcomes).
+
+## Scaled Location Page Rules
+
+- Warning threshold: 30+ location pages.
+- Hard stop threshold: 50+ location pages.
+- Require at least 60% unique content per location page.
+- Require at least 3 localized proof elements per page:
+  - local testimonials/case studies
+  - location-specific service details
+  - local team, office, or coverage details
+
+At hard stop, ask for explicit user justification before proceeding with large-scale rollout.
+
+## Programmatic SEO Gates
+
+Before recommending programmatic generation:
+1. Confirm variable set produces real informational differences.
+2. Confirm pages can include unique utility (data, tools, comparisons, or local proof).
+3. Reject if output would only swap tokens (city/product names) in near-identical copy.
+4. Define canonicalization and indexability strategy up front.
+
+## Duplication Risk Bands
+
+| Similarity Between Pages | Risk Level | Action |
+|--------------------------|------------|--------|
+| < 40% shared text | Low | Proceed |
+| 40% to 60% shared text | Medium | Add differentiators before publishing |
+| > 60% shared text | High | Block rollout until content model is redesigned |
+
+## Schema Content Policy
+
+- Do not recommend HowTo schema as a growth recommendation.
+- Limit FAQ schema recommendations to government and healthcare sites.
+- Ensure structured data mirrors visible page content.
+
+## Gate Outcome Labels
+
+Return one of:
+- `pass`: Meets thresholds
+- `pass_with_risk`: Can ship with documented constraints
+- `fail`: Block and redesign content strategy
+
+Always include concrete remediation steps for `pass_with_risk` and `fail`.
@@ -0,0 +1,61 @@
+# Schema Type Guidance
+
+Prefer JSON-LD. Validate syntax and eligibility before recommending implementation.
+
+## Recommended by Site Type
+
+| Site Type | Primary Schema | Secondary Schema |
+|-----------|----------------|------------------|
+| SaaS | `SoftwareApplication`, `Organization` | `Product`, `BreadcrumbList`, `WebSite` |
+| Local service | `LocalBusiness` (or subtype), `Organization` | `Service`, `Review`, `BreadcrumbList` |
+| E-commerce | `Product`, `Offer`, `AggregateRating` | `Organization`, `BreadcrumbList`, `ItemList` |
+| Publisher | `Article`/`NewsArticle`, `Person`, `Organization` | `BreadcrumbList`, `ImageObject` |
+| Agency | `Organization`, `Service` | `FAQPage` (restricted use), `BreadcrumbList` |
+
+## Important Eligibility Notes
+
+- Do not recommend `HowTo` as an SEO growth tactic.
+- Recommend `FAQPage` only for government and healthcare use cases.
+- Keep markup aligned with visible on-page content.
+- Prefer one coherent graph over fragmented duplicate blocks.
+
+## Global Baseline Types
+
+Implement these where applicable:
+- `Organization` with logo, URL, sameAs
+- `WebSite` with `SearchAction` if site search exists
+- `BreadcrumbList` on hierarchical pages
+
+## E-commerce Essentials
+
+For product pages, include:
+- `name`
+- `description`
+- `image`
+- `sku` or `gtin` when available
+- `brand`
+- `offers` (`price`, `priceCurrency`, `availability`, `url`)
+- `aggregateRating` only when genuine ratings exist
+
+## Publisher Essentials
+
+For article pages, include:
+- `headline`
+- `datePublished`
+- `dateModified`
+- `author`
+- `publisher`
+- `mainEntityOfPage`
+- `image`
+
+## Validation Workflow
+
+1. Detect current schema blocks and type coverage.
+2. Validate syntax and required fields.
+3. Check content parity (structured data must match rendered content).
+4. Remove conflicting or duplicate nodes.
+5. Prioritize schema by impact:
+   - Critical: invalid structured data causing parsing failures
+   - High: missing core schema for primary page type
+   - Medium: missing enrichment fields
+   - Low: optional graph enhancements