semantic-clustering

スター0

フォーク0

更新日2026年5月19日 19:28

Urania's playbook. Topic-cluster authority model, entity SEO, schema.org markup, internal-link wiring, SERP-feature targeting.

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

ソース

lebobo88

lebobo88/RLM-Creative

GitHub リポジトリを開く Creator のリポジトリを見る

ダウンロード

Semantic Clustering

Urania's playbook for organic discovery. Builds topical authority through pillar-cluster architecture, entity SEO, and schema.org markup. Designed for the AI-search era where the corpus is read by LLMs as much as crawled by spiders.

Purpose

Convert CreativeBrief.objective into a defensible organic surface area: pillar pages, cluster pages, internal-link patterns, entity-rich content, and schema markup that earn SERP features and LLM citations.

When to use

A new product or brand needs an SEO foundation.
An existing site has thin or scattered content lacking topical depth.
Polyhymnia's editorial calendar needs SEO-grounded topic selection.
A SERP feature (featured snippet, PAA, knowledge panel) is winnable but unwon.

Inputs

CreativeBrief.objective, target_audience, key_messages
Existing site URL list (sitemap) if available
Keyword research data (search volume, difficulty, intent) — supplied or pulled per tool availability
Competitor SERP analysis for top 3-5 commercial keywords

Outputs

cluster_map.json to RLM/output/seo/{campaign_id}-clusters-{date}.json describing pillars, clusters, intents, internal links
Schema.org JSON-LD snippets per page type
Internal-link patch list for existing pages
SERP-feature target list (snippet, PAA, knowledge panel, video carousel, image pack)

Topic-cluster authority model

Pillar page — a comprehensive, evergreen guide on a head term (e.g. "customer onboarding"). 3000-6000 words. Links out to all its cluster pages and is linked back from each.
Cluster pages — focused articles on long-tail variants ("customer onboarding for SaaS", "customer onboarding KPIs", "customer onboarding software"). 1200-2500 words. Each links to the pillar and to 2-4 sibling clusters.
Hub-and-spoke wiring — pillar is the hub; clusters are spokes. PageRank concentrates on the pillar; topical authority radiates to clusters.
Refresh cadence — pillar reviewed quarterly; clusters reviewed bi-annually; both updated on any product or category change.

One pillar per buyer-journey stage per audience. Do not duplicate pillars across stages (awareness pillar vs consideration pillar vs decision pillar are distinct).

Entity SEO

The modern search engine and the LLM both think in entities, not keywords. For each pillar:

Identify the head entity (the thing the page is about) and 8-15 related entities (people, places, products, concepts) the page should mention.
Cross-reference the entities with Wikidata or Wikipedia where applicable; align spelling and disambiguation.
Mention every related entity at least once with context; link to canonical sources when authoritative.
Build an entity_glossary.md per pillar so future updates preserve coverage.

LLMs cite content that names entities cleanly and provides definitional sentences. Definitional sentences ("X is a Y that does Z") get pulled into AI Overviews and chatbot answers.

Schema.org markup

Minimum schema per page type (JSON-LD in <head>):

Pillar / cluster article — Article or BlogPosting with headline, author, datePublished, dateModified, image, publisher, mainEntityOfPage.
Product page — Product with name, brand, offers (Offer with price, priceCurrency, availability), aggregateRating if applicable, review if applicable.
FAQ section — FAQPage with mainEntity array of Question -> acceptedAnswer.
How-to — HowTo with step array.
Video — VideoObject with name, description, thumbnailUrl, uploadDate, duration (ISO 8601), contentUrl.
Organization (site-wide) — Organization on the home page with logo, sameAs (social profiles), contactPoint.
Breadcrumb — BreadcrumbList on every non-home page.

Validate every snippet with Google's Rich Results Test before publish.

Internal-link patterns

Every cluster page links to its pillar with descriptive anchor text containing the pillar's head term.
Pillar page links out to all its clusters in a single "Explore the topic" section AND in-line where contextually relevant.
Each cluster links to 2-4 sibling clusters in-line, not just at the bottom.
Older posts get retro-fitted links to new pillars/clusters (this is the highest-leverage SEO maintenance task).
Avoid nofollow on internal links unless gating UGC.
Use 60-80% descriptive anchor text; 20-40% generic ("read more", "learn how") for naturalness.

SERP-feature targeting

Featured snippet (paragraph) — open the relevant section with a 40-60 word direct answer to the query in plain prose. Use the query phrasing in the H2 above it.
Featured snippet (list) — H2 in the form "How to X" or "X steps to Y"; follow with an ordered or unordered list of 5-8 items.
Featured snippet (table) — H2 with comparative intent; follow with a markdown / HTML table of 3-5 columns.
People Also Ask — include a FAQ section with FAQPage schema; phrase questions exactly as users ask them.
Knowledge panel — Organization schema, consistent NAP (name/address/phone), populated Wikidata entry where eligible, brand mentions across authoritative outlets.
Image pack — descriptive alt text, <figure> + <figcaption>, file names that match topic, image sitemap entry.
Video carousel — VideoObject schema, transcript on page, hosted on YouTube + embedded, chapters via Clip schema.
Top stories — requires NewsArticle schema + Google News inclusion; news-cadence sites only.
LLM citation (AI Overviews, Perplexity, ChatGPT browse) — clean entity coverage, definitional sentences, freshness signals (dateModified), authoritative outbound citations, named author with bio page.

Procedure

Identify 3-5 head terms most aligned with objective and audience.
For each head term, propose a pillar page; map 5-10 cluster pages per pillar with intent classification (informational / commercial / transactional / navigational).
Build entity lists per pillar; cross-check with Wikidata.
Audit existing site content; map old URLs to new clusters; plan redirects for retired pages (301 only).
Draft schema.org JSON-LD per page type; test with Rich Results Test.
Draft internal-link patch list: which existing pages must link to which new pillars/clusters.
Coordinate publication cadence with Polyhymnia's editorial-calendar so SEO content is also part of the content arc.
After publish: monitor SERP-feature acquisition, impressions, clicks, position; log to eights.memory.remember(tags=["seo","cluster","serp"]); quarterly refresh.

Validation checklist

Every pillar has 3-10 cluster pages mapped
Every cluster page links to its pillar AND 2-4 siblings
Every page has schema.org markup appropriate to its type, validated
Every pillar has an entity glossary documented
No two pages target the same primary keyword (cannibalization check)
All redirects are 301; no chains; no loops
FAQ schema used only where FAQs are visibly on the page (Google penalty risk otherwise)

Failure modes / escalation

Cannibalization detected — consolidate via 301; pick the stronger URL; do not run two competing pages.
Brand penalty / manual action — pause SEO production; route to Calliope and brand-safety; remediate before continuing.
Schema validation fails — fix before publish; rich results lost is worse than no rich results attempted.
Algorithm update tanks rankings — wait 2-4 weeks before structural response; document hypothesis; recall similar past episodes.

References

Editorial cadence: editorial-calendar (Polyhymnia)
Copy execution: platform-voice (Erato + Terpsichore)
Memory tags: seo, cluster, pillar, serp, entity

name	semantic-clustering
description	Urania's playbook. Topic-cluster authority model, entity SEO, schema.org markup, internal-link wiring, SERP-feature targeting.
user-invocable	false
argument-hint	<cluster\|entity\|schema\|interlink>
allowed-tools	["Read","Write"]

name	semantic-clustering
description	Urania's playbook. Topic-cluster authority model, entity SEO, schema.org markup, internal-link wiring, SERP-feature targeting.
user-invocable	false
argument-hint	<cluster\|entity\|schema\|interlink>
allowed-tools	["Read","Write"]

semantic-clustering

このリポジトリの他の Skills

Semantic Clustering

Purpose

When to use

Inputs

Outputs

Topic-cluster authority model

Entity SEO

Schema.org markup

Internal-link patterns

SERP-feature targeting

Procedure

Validation checklist

Failure modes / escalation

References

Semantic Clustering

Purpose

When to use

Inputs

Outputs

Topic-cluster authority model

Entity SEO

Schema.org markup

Internal-link patterns

SERP-feature targeting

Procedure

Validation checklist

Failure modes / escalation

References

このリポジトリの他の Skills