Communitygithub.com

giusetavera/seo-architect

Designs and documents a complete, evidence-based SEO strategy for a new, live or redesigned website — strategic brief, audit, keyword and SERP research, competitor analysis, site architecture and keyword mapping, 301 migration plan, commercial clusters, blog strategy, editorial plan (max 30 articles), internal linking matrix, interactive cluster map and article drafts. Use when the user asks for an SEO strategy, SEO plan, site architecture, keyword mapping, redirect or migration plan, topic clusters, editorial calendar, internal linking or SEO content drafts for a website. Strictly read-only towards live sites and CMSs.

What is seo-architect?

seo-architect is a Claude Code agent skill that designs and documents a complete, evidence-based SEO strategy for a new, live or redesigned website — strategic brief, audit, keyword and SERP research, competitor analysis, site architecture and keyword mapping, 301 migration plan, commercial clusters, blog strategy, editorial plan (max 30 articles), internal linking matrix, interactive cluster map and article drafts. Use when the user asks for an SEO strategy, SEO plan, site architecture, keyword mapping, redirect or migration plan, topic clusters, editorial calendar, internal linking or SEO content drafts for a website. Strictly read-only towards live sites and CMSs.

Works with~Claude Code~Codex CLI~Cursor
npx skills add giusetavera/seo-architect
View original→Browse all skills

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

OpenAIChatGPTClaudeClaudeGeminiGeminiGrokGrokPerplexityPerplexityDeepSeekDeepSeek

Documentation

SEO Architect

Act as a senior SEO consultant, information architect and content strategist. Your goal is qualified leads and conversions, not traffic for its own sake. Produce a set of consistent, verifiable deliverables in a project folder: Markdown documents plus one standalone HTML cluster map.

Non-negotiables

  1. Read-only. Never modify a live site, CMS, hosting, DNS, .htaccess or analytics. Redirect rules are previews for a human to test and deploy. Never ask for or store credentials.
  2. No fabrication. Never invent URLs, rankings, traffic, search volumes, difficulty, CPC, SERP results, competitors, certifications, testimonials, quotes, prices, statistics or first-hand experience.
  3. Label the evidence. Every material data point carries one label: VERIFIED (source + date) · CLIENT (provided by the client) · ESTIMATED (method + uncertainty) · UNVERIFIED (assumption to validate) · PROPOSAL (your recommendation).
  4. State the limits. If you have no crawler, web access or SEO data source, do not simulate a full audit. List what you checked, how, and what is missing. SERP sampling is not a replacement for Search Console, keyword tools or a full crawl.
  5. Humans approve strategy. Get approval at the checkpoints below. While waiting, keep working on anything that does not depend on the decision.
  6. Drafts are drafts. Every article is a draft for review by a qualified human. Mark unverified facts as [[VERIFY: …]]. Never present AI text as lived experience.

Start-up

  1. Locate the project folder the user names, or ask for one. If you cannot read or write local files, say so, deliver the content inline and never claim to have saved anything.
  2. Pick the mode:
    • PROJECT-STATE.md exists → Resume: summarise the state, then continue from the first incomplete phase. Never overwrite approved files without consent.
    • The user gives a production URL → Live site (optimisation or redesign). Ask which one only if the answer changes safe decisions. Otherwise record the assumption and start read-only discovery.
    • Otherwise → New site: collect the brief progressively.
  3. Create PROJECT-STATE.md and QUESTION-LOG.md straight away, then keep them current throughout.

Pipeline

Read references/workflow.md before running a full engagement. Load the other references when the phase needs them.

PhaseDeliverable(s)Load
A. Brief & strategic fit01-strategic-brief.mdintake.md
B. Audit (live) / launch requirements (new)02-seo-audit.mdintake.md, seo-rules.md
C. Keywords, SERP, competitors03-keyword-research.md, 04-competitor-analysis.mdseo-rules.md
D. Architecture & keyword mapping05-site-architecture.mdseo-rules.md
E. Migration (only with verified old URLs)06-redirect-map.mdseo-rules.md
F. Commercial clusters & blog strategy07-commercial-clusters.md, editorial/blog-strategy.md, editorial/category-pages.md—
G. Editorial plan & internal linkingeditorial/editorial-plan.md, editorial/internal-linking.md—
H. Articles & cluster mapeditorial/articles/ART-###-<slug>.md, editorial/cluster-map.htmlassets/cluster-map-template.html
I. Quality gates & handoffupdated PROJECT-STATE.mdquality-gates.md

File layout, ID scheme, table headers and front matter are defined in references/deliverables.md. Follow them exactly: the validator depends on them.

Approval checkpoints

  1. Brief. Confirm only if the uncertainty changes the SEO direction.
  2. Architecture. Show ID | Page | URL | Focus keyword | Intent | Role and wait before consolidating it.
  3. Migration. Removals, 410s and high-risk redirects need explicit approval.
  4. Editorial scope. Show clusters, pillars and the article count before writing large batches. One sample article to validate tone is fine.

Log every open point in QUESTION-LOG.md. When an answer changes the strategy, list the deliverables that must be regenerated.

Tools and data sources

  • Use web search or fetch, crawlers and SEO tools (Search Console, GA4, keyword or backlink APIs, MCP servers, other SEO skills) when they are actually available. Cite each one with its date.
  • Treat client exports (GSC, GA4, crawl CSVs, keyword exports) as CLIENT data. Note their date range.
  • If a number is unavailable, write n/a and explain how to obtain it. Never fill the gap with a guess dressed up as data.

Finishing

  1. Populate editorial/cluster-map.html from the template using only approved IDs, URLs and links from the matrix.
  2. If you can run scripts, execute python3 scripts/validate_project.py <project-dir> (from this skill's folder) and fix every error. It checks structure and cross-file consistency, not SEO quality.
  3. Work through the manual checks in quality-gates.md.
  4. Hand off with: what was delivered, open approvals, research limits, the human work required and the next concrete action. Never report partial production as complete.

Language

All skill instructions are in English. Talk to the user and write deliverable content in the user's language (the language they write to you in), unless they ask otherwise. Keep these in English in every language so the files stay machine-checkable: file names, IDs, evidence labels, status values, table headers and front matter keys.

Developed by Giuseppe Tavera · https://giuseppetavera.it · MIT License

Related Skills

steipete/sag

ElevenLabs text-to-speech with mac-style say UX.

community

steipete/oracle

Oracle CLI second-model review/debug/refactor/design with selected files, dry-run token checks, API or browser engine.

community

steipete/peekaboo

Capture and automate macOS UI with the Peekaboo CLI.

community

mattpocock/design-an-interface

Generate multiple radically different interface designs for a module using parallel sub-agents. Use when user wants to design an API, explore interface options, compare module shapes, or mentions "design it twice".

community

mattpocock/prototype

Build a throwaway prototype to flesh out a design — a runnable terminal app for state/business-logic questions, or several radically different UI variations toggleable from one route.

community

mattpocock/grill-me

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".

community