SEOAgent

A persistent, repo-local SEO workflow. Unlike one-off SEO prompts that lose their context the moment the conversation ends, every artifact here is written to a .seoagent/ directory in the repo — so audits, strategy, briefs, and drafts accumulate across sessions instead of resetting each time.

When to Use This Skill

Use this skill when the user wants to:

• Run a technical SEO audit of a site (meta tags, headings, internal links, schema, Core Web Vitals readiness, AI-search optimization)
• Build a keyword strategy with topic clusters and an internal-linking structure
• Write a content brief for a specific page or article
• Draft an SEO-optimized article with full metadata and JSON-LD
• Resume SEO work — read back prior audit findings, strategy, or the roadmap

Workspace Model

All work persists to .seoagent/ so it survives across sessions:

.seoagent/
  project.md           # domain, site type
  context.md           # business context, tone, banned topics
  audit/latest.md      # findings as [ ] / [x] checkboxes, tagged by severity
  strategy/clusters/   # hub-and-spoke topic clusters
  briefs/{slug}.md     # page-type-aware briefs
  content/{slug}.md    # drafts with SEO frontmatter + JSON-LD
  roadmap.md           # prioritized next steps
  changelog.md         # history of SEO work

Always read existing .seoagent/ files before acting, and write results back so the next session can build on them.

Workflow (5 phases)

1. Technical audit

Fetch the site's key pages and check each against the list below. Save findings to .seoagent/audit/latest.md as [ ] checkboxes the user can flip to [x] when fixed, each tagged critical | high | medium | low.

• Indexability: no accidental noindex, no blocked paths in robots.txt, canonical present and self-referential
• Title tag: present, unique, 50–60 chars, primary keyword near the front
• Meta description: present, unique, 140–160 chars, compelling
• One <h1> per page, descriptive; logical h2/h3 hierarchy
• Internal links: no orphan pages; descriptive anchor text
• Structured data: appropriate JSON-LD (Organization/Article/Product/FAQ) present and valid

• Image alt text on content images
• OpenGraph + Twitter card tags for social sharing
• Core Web Vitals readiness (LCP image preloaded, no layout shift, reasonable JS)
• Clean, keyword-relevant URL slugs
• XML sitemap present and submitted; HTTPS enforced

2. Keyword strategy (hub-and-spoke)

Research the niche, then build topic clusters where each cluster has roles:

• PILLAR — broad, high-value hub page
• SUB_PILLAR — focused subtopics linking up to the pillar
• LONG_TAIL — specific questions/niche queries linking up to sub-pillars

Internal links funnel authority up toward pillars. Save clusters to .seoagent/strategy/clusters/{slug}.md with the article table + link graph.

3. Content briefs (page-type-aware)

Different page types need different structures. Pick the protocol by type:

• landing — conversion-focused, Product/Service JSON-LD
• pillar — comprehensive overview, links to all sub-pillars
• sub_pillar — focused subtopic depth
• long_tail — direct answer to a specific query, FAQPage JSON-LD
• programmatic — templated pages from a data set

Each brief gets a URL pattern, section outline (H2/H3), internal-link plan, JSON-LD plan, and a word-count target. Save to .seoagent/briefs/{slug}.md.

4. Draft articles

Write from the brief with complete SEO frontmatter:

• meta_title, meta_description, canonical
• OpenGraph + Twitter fields
• JSON-LD: Article, plus FAQPage / HowTo where the content warrants it
• An image plan (hero + inline) with alt text

Save to .seoagent/content/{slug}.md.

5. Monitor

Re-audit periodically, update .seoagent/roadmap.md with the next highest-leverage actions, and append what changed to .seoagent/changelog.md.

Answer-engine optimization (AEO/GEO)

Beyond classic SEO, structure content so AI assistants (ChatGPT, Claude, Perplexity, Google AI Overviews) can cite it: lead with a direct answer, use clear headings as questions, include comparison tables and concrete stats, and add FAQ/HowTo structured data. This makes pages citable in AI-generated answers, not just rankable.

Running the full agent

This skill captures the SEOAgent methodology. The full agent ships as a CLI that scaffolds the .seoagent/ workspace, installs an expanded reference library (per-page-type protocols, a JSON-LD library, an audit-check catalog), and adds an optional cloud loop:

# npm
npm install -g @seoagent-official/seoagent && seoagent init

# or the plugin marketplace (Claude Code or Codex)
/plugin marketplace add Baxter-Inc/seoagent-npm

Free and local by default (uses your agent's own tools; data stays in your repo). Optional cloud adds real keyword data, Google Search Console, and CMS publishing. MIT licensed — https://github.com/Baxter-Inc/seoagent-npm