claude-seo (Claude Code universal SEO agent workflow)

I came across a new Claude Code workflow called claude-seo that turns “do SEO” into something more like an engineer-to-engineer build: multiple specialized sub-skills and sub-agents, plus optional extensions for pulling live SEO data.

Where it’s useful is when you’re asked to ship an SEO deliverable for a real codebase: you inherit a site, need a technical audit (indexing, crawlability, schema coverage, Lighthouse-style agentic browsing), and then you need consistent outputs (reports for technical findings and supporting artifacts like schema checks, PDF/Excel exports, and structured documentation).

What to look at first is the entry point that lists the 26 sub-skills and 19 sub-agents—specifically the agent readiness pieces (e.g., Lighthouse agentic browsing, WebMCP, and the llms.txt-style readiness signal), because those are the “make the agent actually navigate and reason” layers rather than the purely content-focused parts.

The workflow scope is broad on purpose (international/local/e-commerce SEO, backlinks, Google APIs, GEO/AEO), but the practical value is that it organizes those tasks into reusable modules, so you can run only the parts you need for a given engagement.

If you’re evaluating it for production-ish use, I’d focus on: - How the sub-agents are orchestrated (boundaries between discovery, analysis, and reporting) - Whether the workflow supports deterministic inputs/outputs suitable for CI-like runs - How optional live-data extensions are gated so offline runs still work

Overall: it’s a shipping-oriented example of “SEO as an agent system,” rather than a generic prompt pack.