I built 4 Claude skills that run a whole SEO blog pipeline - topic research through published post

Hey PH - been running content automation for a few months and figured I'd share what actually holds up.

Here's the thing about AI-written blogs: drafting is the easy part. Posts that rank need topic validation, internal links, JSON-LD schema, and an AI-SEO check before publish. Handle each of those manually and you've just traded one bottleneck for four. So I turned the pipeline into four Claude skills that chain together.

Topic Research pulls from People Also Ask, Reddit, Stack Overflow, and GitHub issues before you commit to anything. The goal is avoiding that situation where you finish a 2,000-word draft and realize nobody's searching for it.

SEO Blog Writer handles the rest of pre-publish: classification, research, HTML draft, LLM scrub, AI-SEO audit. Has adapters for Ghost, WordPress, Webflow, Sanity, and static site generators.

Blog Figure SVG I added because I kept skipping illustrations. It generates diagrams and feature cards as SVG with screen-reader metadata, which is good enough that I actually use them.

Editorial Calendar maintains a scored topic backlog and slots posts into a rolling schedule. When the backlog empties it refills from the research skill - I haven't had to manually queue a topic in weeks.

npx skills add AutomateLab-tech/publishing-skills

Claude Code, Cursor, any Claude-backed agent. MIT-0.

What parts of a publishing pipeline have you found genuinely hard to automate?

GitHub: https://github.com/AutomateLab-tech/publishing-skills