Skills, not one giant prompt
Each concern gets its own skill with its own instructions and tools, so a technical crawl, a schema check and a content review stay independently useful and independently testable.
An open-source suite of 32 SEO skills for Claude Code that runs a full search programme end to end, from technical audits to AI-search visibility.

ProSEO turns SEO from a pile of one-off checks into a staged programme an AI agent can actually run. Built on Anthropic's Claude Agent SDK, it ships 32 skills, 18 sub-agents that work in parallel and 53 Python tools, organised into a seven-stage lifecycle that carries a site from discovery through to measurement and iteration.
SEO tooling is fragmented. Crawlers, keyword tools, schema validators and rank trackers each solve a slice, and nobody stitches the slices into a repeatable programme, so the same audit gets improvised from scratch every time.
Agents make that worse before they make it better: a single prompt cannot hold a technical audit, content review, local presence and AI-search visibility at once without losing the thread.
Each concern gets its own skill with its own instructions and tools, so a technical crawl, a schema check and a content review stay independently useful and independently testable.
Eighteen specialist sub-agents fan out across a site at once, each returning structured findings, which keeps a full audit within one context instead of serialising every check.
Seven stages, Discover through Iterate, track progress between sessions so a programme resumes where it left off rather than restarting from an empty page.
I build AI-native SaaS, healthcare products, marketplaces and internal platforms from first commit to launch.