I don't see any clear evidence in this article that blogpost and PR was opened by openclaw agent and not simply by human puppeteer. How can the author know that PR was opened by agent and not by human? It is certainly possible someone set up this agent, and it's probably not that complex to set it up to simply create PR, react to merge/reject on blogposts, but how does author know this is what happened?