wtf is real-agent-workflows?

Real Agent Workflows is a collection of AI automation setups that the author actually runs on their own computer, shared one at a time rather than presented as a single polished product. Each folder in the repository is a complete workflow: what triggers it, the steps it takes, and what it produces, written up as a plain markdown playbook called SKILL.md. The author is explicit that this is a guide, not a plugin, and that it is not tied to any particular AI tool, capability names like web search or browser control in the instructions are meant to be mapped to whatever equivalent tool your own AI assistant provides. The repository draws a distinction between a skill, a single capability used on demand, and a workflow, the full chain of trigger, steps, and output that runs itself on a schedule or event. Six workflows are marked ready today, covering things like turning the last day of AI related news into a categorized briefing inside an Obsidian notes vault, pulling several news sources plus an email inbox into one page you can approve with checkboxes each morning, harvesting and filtering saved social media posts into knowledge cards, filing any pasted link as a summarized note, drafting a short video script from a topic in your notes, and triaging a day's email into an action required or for your information list. Many more workflows, covering things like website building, search engine optimization, and video editing, are listed as still in progress. To use a workflow, you copy its SKILL.md file into your own AI agent's skills folder, run it once by hand to check the result, then put it on a recurring schedule if it is useful. The author suggests trying just one workflow closest to your actual work for a couple of weeks before deciding whether to keep it. Most of the individual workflow documentation is currently written in Chinese, though the top level repository itself is bilingual. The project is released under the MIT license.