unclejobs-ai/blog-posting-auto

blog-posting-auto is a Korean-language blog writing pipeline built on top of Claude Code, the command-line coding agent. You give it a single topic line, and a chain of six small agents takes over: one researches the topic, one fact-checks what the researcher found, one plans the article, one writes the draft, one reviews and polishes it, and the last one pushes the result into a Notion database as a Draft page. The author says the part that sets this apart from other blog automations is the fact-checker step. The researcher gathers raw claims, and the fact-checker re-visits the original source URLs to see if each claim really holds up. Only claims that pass survive into a verified file, and the writer is only allowed to read that verified file. The aim is to stop hallucinations and fake first-person stories from sneaking into the final post. To run it you need Claude Code installed, the Notion connector set up as a global MCP server, and one Notion database to drop posts into. Setup is a git clone, a copy of the example env file, and filling in the Notion parent page ID and database ID. Then from inside a Claude Code session you call a slash command like /blog-pipeline followed by your topic. The pipeline assigns different Claude models to different jobs, with the cheaper Haiku model for routine search and posting calls, Sonnet for fact-checking, planning, and writing, and Opus for the final review. Style rules, banned phrases, and a checklist live in a separate folder so you can swap them out for English or technical writing. Publishing to Notion always asks for your explicit OK first.