Growth seo orchestrator

Growth-SEO Orchestrator

The single entry skill for SEO and search-growth work. It locates the task on the funnel stage × concern map and delegates to one of 8 specialist spokes. The model every SEO task shares — search intent, the query→page contract, and the E-E-A-T/quality bar — lives in growth-seo-core; read it before mapping keywords to pages or judging whether content will rank.

Routing map (intent → spoke)

Understand the rules

• "What does Google actually want?" / first principles, Search Console, crawl/index basics → google-official-seo-guide

Research demand & intent

• Niche / offer / topic → keyword map + audience-question map + an executable research pipeline → searchintentautomation
• Turn that demand into a structured, writer-ready brief → seo-content-brief

Plan the site

• Page hierarchy, navigation, URL structure, internal linking, "what pages do I need" → site-architecture

Produce & enrich content

• Add/fix structured data, JSON-LD, rich-result eligibility → schema-markup
• Build many keyword/location/comparison pages from templates + data → programmatic-seo

Diagnose & optimize

• Audit, "why am I not ranking," technical/on-page issues, traffic drop, Core Web Vitals → seo-audit
• Get cited by LLMs / appear in AI Overviews, ChatGPT, Perplexity (AEO/GEO/LLMO) → ai-seo

Picked-up spokes

Granular execution specialists folded in from the claude-seo library (MIT). The 8 spokes above stay the primary map; route to these when the task is narrower or needs live data. They complement, not replace, the spokes above (e.g. seo-audit stays the entry for "why am I not ranking"; drop to seo-technical/seo-page for a deep crawl/single-page pass).

Diagnose at finer grain

• One URL, full on-page + content + schema + image + meta pass with a score card → seo-page
• Infrastructure audit — crawlability, indexability, security headers, Core Web Vitals, JS rendering, robots.txt / AI-crawler access, IndexNow → seo-technical
• Content-only review — E-E-A-T, readability, thin-content risk, AI-citation readiness → seo-content
• Image SEO + CWV — alt text, WebP/AVIF, srcset, lazy loading, CLS → seo-images

Build & validate artifacts

• Competitor comparison / "X vs Y" / alternatives / roundup pages with schema → seo-competitor-pages
• Generate or validate an XML sitemap (format, limits, lastmod, canonical hygiene) → seo-sitemap
• Hreflang / international SEO — validate or generate locale alternates → seo-hreflang

Plan & pull live data

• Full strategic SEO plan/roadmap with industry templates, competitive analysis, KPIs, phasing → seo-plan
• Live SEO data (real SERPs, keyword volumes/difficulty/intent, backlinks, on-page, LLM mentions) via the DataForSEO MCP → seo-dataforseo (feeds every other spoke when available)

Standard Operating Flow

1. Locate the task: which funnel stage (understand → research → architect → produce → optimize) and which concern.
2. If it touches what query a page targets, intent match, or the quality/E-E-A-T bar, pull the model from growth-seo-core first — these decisions cascade into every later step.
3. Delegate to the spoke(s). Multi-step asks fan out in funnel order (e.g. "rank a new section" → searchintentautomation → seo-content-brief → site-architecture → programmatic-seo/content → schema-markup → seo-audit).
4. Return: chosen spoke(s), the target query/intent and page it maps to, the success signal (ranking, rich result, AI citation, traffic), and the next action.

Guardrails

See growth-seo-core. In short: intent before output — never produce a page without naming the query it serves and the intent behind it; one primary query per page (don't cannibalize); match content type to intent (informational vs commercial vs transactional). Earn rankings with genuine E-E-A-T, never manipulate; programmatic pages must add real value per URL, not thin doorways; and AI-search visibility is downstream of the same intent + authority, not a separate trick. State any change that affects what a URL targets or how it's indexed.

Loading spokes on demand

To keep CLI startup context lean, this cluster's spokes are not separately registered as skills — only this orchestrator and its *-core are enumerated. When you route to a spoke named above, load it on demand by reading its file:

~/.agents/skill-clusters/skills/<spoke-name>/SKILL.md (or skills/<spoke-name>/SKILL.md inside the skill-clusters repo).