| name | amw-seo |
| description | SEO + Core Web Vitals evaluation framework (E-E-A-T quality signals, LCP/INP/CLS, structured data, technical foundations). Triggers on "SEO", "core web vitals", "E-E-A-T", "structured data", "schema markup", "page performance ranking", "search ranking", NOT generic design intent. Use when evaluating or improving a webpage's SEO, Core Web Vitals, or structured data markup. Trigger with explicit "SEO", "core web vitals", or "structured data" phrasing. |
SEO Evaluation Reference
Orchestrated by: SKILL.
Overview
SEO evaluation and Core Web Vitals interpretation framework. Knowledge-base only — interprets measurements from amw-dev-browser or user-supplied data; does not capture metrics itself. Covers 8 evaluation areas: E-E-A-T quality signals, Core Web Vitals (LCP/INP/CLS), technical SEO, content SEO, structured data / JSON-LD, AI-assisted content principles, relative importance of factors, and multi-signal measurement. Pairs with amw-ux-evaluator for combined UX + SEO scoring.
Instructions
- Classify the user's SEO concern into one of the 8 evaluation areas (E-E-A-T, Core Web Vitals, Technical SEO, Content SEO, Structured Data, AI-assisted content, Relative importance, Measurement).
- Walk the
## Technique selection tree top-down and open only the single TECH reference file whose TOC matches the current concern.
- Gather live metrics when needed via SKILL (LCP candidate identification, resource sizes, layout shift observation, interaction latency); interpret them through the framework — this skill does not capture metrics itself.
- Score against the relevant thresholds (LCP < 2.5s, INP < 200ms, CLS < 0.1 for Core Web Vitals; E-E-A-T signals for content quality) and produce a prioritized list of findings.
- When a combined UX + SEO audit is requested, pair with SKILL — run both evaluations and merge the finding lists by priority.
Walk the ## Technique selection tree to pick the matching TECH reference for the user's SEO concern. Read only the file whose TOC matches the current need.
Activation
No dedicated slash command — this skill has no matching /amw-* shortcut. Invoked by the design-principles orchestrator during Phase B when the user requests SEO scoring alongside design work, or when /amw-eval triggers an SEO + UX combined audit. Also callable directly when the user explicitly names SEO, Core Web Vitals, E-E-A-T, or structured data.
This skill is autonomous and self-contained — any agent (the main-agent, a sub-agent, or an external orchestrator) can use it by reading this SKILL.md and its references. The skill's techniques are NOT limited to what matching commands expose.
Position in flow
VALIDATION / REFERENCE. Evaluation framework used alongside ux-evaluator when the user wants SEO + design quality scored together. Knowledge-base only — interprets measurements; does not capture them.
Trigger conditions
- "audit SEO", "SEO review", "SEO score"
- "core web vitals", "LCP", "INP", "CLS", "page experience"
- "E-E-A-T", "experience expertise authoritativeness trust"
- "structured data", "schema markup", "JSON-LD", "rich results"
- "search ranking factors", "why does this page rank"
- "is this page SEO-ready"
Does NOT trigger on: generic "make this look better", pure visual critique, copywriting requests without search-intent framing.
Prerequisites
- runtime_binaries: none (framework is prose-encoded)
- Pairs with SKILL for live capture: LCP candidate identification, resource sizes, layout shift observation, interaction latency
- Pairs with SKILL when user requests combined UX + SEO scoring
Evaluation framework sections
- E-E-A-T — Experience, Expertise, Authoritativeness, Trust. Quality lens, not a direct ranking factor; weighted heavily on YMYL topics.
- Core Web Vitals — LCP < 2.5s, INP < 200ms, CLS < 0.1. Rarely override poor content but hold back otherwise-good pages.
- Technical SEO — Crawl/index control (sitemap, robots.txt, canonical, status codes, HTTPS), performance, semantic HTML.
- Content SEO — Page elements (title, meta, H1, headings, alt) + quality signals (depth, originality, accuracy, clarity, usefulness).
- Structured data — Schema.org / JSON-LD types enable rich-result eligibility; do not boost rankings directly.
- AI-assisted content — Output quality matters, not authorship method. Unedited AI output with factual errors or thin value is the failure mode.
- Relative importance — No fixed order. When pages compete closely: content quality > authority/trust > page experience > mobile > technical accessibility.
- Measurement — Multi-signal validation: visibility, engagement, performance (CWV field data), coverage, authority.
Technique selection
Walk this decision tree top-down to pick the right reference. If a branch does not match the user's intent, skip to the next. Every technique in the catalog is a leaf of this tree.
- Is the user asking about this skill's domain?
[TECH-seo-ai-content.md] What it does · When to use · How it works · Effective use · Risky use · Minimal example · Gotchas · Cross-references
[TECH-seo-content-quality.md] What it does · When to use · How it works · Page-level elements · Content quality signals · Intent classification · Minimal example · Gotchas · Cross-references
- For "Core Web Vitals — page-experience signals" -> TECH-seo-cwv
[TECH-seo-cwv.md] What it does · When to use · How it works · Field data vs lab data · Important context · Minimal example · Gotchas · Cross-references
[TECH-seo-measurement.md] What it does · When to use · How it works · Cross-validation examples · KPIs worth tracking vs vanity metrics · Minimal example · Gotchas · Cross-references
[TECH-seo-relative-importance.md] What it does · When to use · How it works · Approximate weight hierarchy · The operational rule · Remediation priority · Minimal example · Gotchas · Cross-references
[TECH-seo-technical.md] What it does · When to use · How it works · Crawl & index control · Performance & accessibility (technical prerequisites for CWV) · Common technical failures · Minimal example · Gotchas · Cross-references
References
Every technique in this skill is documented as a single reference file under ./references/. Each file follows the standard TOC: What it does · When to use · How it works · Minimal example · Gotchas · Cross-references. The technique-selection tree above is the authoritative router — open only the file whose tree branch matches the current need.
[TECH-seo-ai-content.md] What it does · When to use · How it works · Effective use · Risky use · Minimal example · Gotchas · Cross-references
[TECH-seo-content-quality.md] What it does · When to use · How it works · Page-level elements · Content quality signals · Intent classification · Minimal example · Gotchas · Cross-references
[TECH-seo-cwv.md] What it does · When to use · How it works · Field data vs lab data · Important context · Minimal example · Gotchas · Cross-references
[TECH-seo-eeat.md] What it does · When to use · How it works · Minimal example · Gotchas · Cross-references
[TECH-seo-measurement.md] What it does · When to use · How it works · Cross-validation examples · KPIs worth tracking vs vanity metrics · Minimal example · Gotchas · Cross-references
[TECH-seo-relative-importance.md] What it does · When to use · How it works · Approximate weight hierarchy · The operational rule · Remediation priority · Minimal example · Gotchas · Cross-references
[TECH-seo-structured-data.md] What it does · When to use · How it works · Minimal example · Gotchas · Cross-references
[TECH-seo-technical.md] What it does · When to use · How it works · Crawl & index control · Performance & accessibility (technical prerequisites for CWV) · Common technical failures · Minimal example · Gotchas · Cross-references
Completion checklist
Before reporting a job using this skill as complete, verify every item below. FAIL on any item should trigger a remediation loop; do not deliver partial work.
- Inputs captured verbatim from the user (brief, URL, reference files) — no silent paraphrasing that changes meaning.
- At least one
TECH-*.md file from skills/amw-seo/references/ was consulted and is cited in the final report.
- Output passes the skill's own non-negotiables (see the
Non-negotiables section below if present).
- No AI-slop per ai-slop-avoid (generic gradients, stock-photo hero, fake testimonials, lorem copy, CTA-hero-features-testimonials template).
[ai-slop-avoid.md] I. Visual style · II. Typography · III. Layout · IV. Content and copy · V. Interaction and motion · VI. Color · Self-check workflow · VII. Content density principle (positive stance) · VIII. Content anti-patterns (T-042) · IX. Anti-AI-cliché visual checklist (T-044) · X. Production-test tells (taste-skill, MIT)
- If the skill emits HTML/SVG/ASCII, the output was rendered/validated by the matching tool (
bin/amw-validate-ascii.py, bin/amw-html-export.py, bin/amw-svg-render.py, etc.).
- Cross-skill hand-offs documented — if work routed through another skill, that skill's SKILL.md + TECH file are named in the report.
- User-facing filename is descriptive English (
Login Flow.html, not output.html).
Output
This skill produces TWO kinds of output:
- Artifact(s) — the actual work product (e.g. SEO evaluation report
.md files; interpretation only — no measurements taken). The output path is determined by project inference, NOT hardcoded. See project-output-routing for the full detection rules. Summary of the priority order:
[project-output-routing.md] When to consult this doc · Detection order · Per-artifact-type default subpath · Reconciliation when multiple candidates match · Edge cases · Quick-reference algorithm (pseudo-code) · Cross-references
- User-supplied path (honor verbatim)
- Framework convention (React/Vite/Next/Astro →
./src/...; Flutter → ./lib/; etc.)
- Existing
./design/<subtype>/ folder if present
- Generic fallback (
./design/references/ or ./reports/webdesigner/ created fresh)
- Last-resort scratch:
/tmp/amw-seo-<slug>/
Every artifact file is listed with its path in the report (next item).
-
Job-completion report — a markdown file at:
$MAIN_ROOT/reports/webdesigner/<YYYYMMDD_HHMMSS±HHMM>_<title-slug>_<8-char-hash>.md
The report must contain, in order:
- Inputs — what the user provided + any auto-detected context
- Method — which TECH references were consulted, which pipeline steps ran
- Artifacts — bullet list, one per produced file, formatted as:
- <artifact-path> — <1-line description> — **How to use:** <usage tip> — **Next steps:** <suggested follow-up>
- Checklist — each item from the Completion checklist above, with PASS / FAIL / N/A
- Deviations — any step skipped or changed, with rationale
The <8-char-hash> is a short content-addressed hash of the report body (e.g. first 8 chars of SHA-256 of the inputs+artifacts list) for uniqueness.
Resolve $MAIN_ROOT via git worktree list | head -n1 | awk '{print $1}'.
Every artifact MUST be linked from the report. reports/webdesigner/ is for user-facing job outputs; reports/audit/ is for build-time audits — keep them separate.
Examples
See each TECH file under ./references/ for the "Minimal example" section. E.g. TECH-seo-cwv.md shows a PageSpeed Insights interpretation workflow; TECH-seo-structured-data.md shows a JSON-LD Article schema snippet.
Resources
- SKILL — orchestrator
- SKILL — live page inspection for CWV capture
- SKILL — pairs for combined UX + SEO scoring
/amw-eval — user-facing command that runs ux-evaluator + seo together
Non-negotiables
- Does NOT claim generic design / UI work.
- Knowledge-base only — does NOT run measurements itself; pairs with
dev-browser for live capture.
- Does NOT substitute for Lighthouse / PageSpeed Insights / CrUX — those measure; this skill interprets.
- Does NOT promise ranking outcomes. SEO is probabilistic; factor weights shift over time.
Error Handling
- JS-heavy SPA that does not render without JS → CWV capture misleading; flag and request SSR / pre-rendered variant before scoring.
- Auth wall / paywalled page → cannot observe what crawlers see; request public-equivalent URL or admit evaluation is partial.
- Offline environment → no live CWV, no structured-data validator access; restrict output to static-analysis findings and mark dynamic claims as unverified.
- User asks for guaranteed ranking lift → decline; SEO evaluation reports quality signals, not predictions.
