Post

Recently, there was some controversy, because an AI user pointed their automated tools at our website, and got a warning about a prompt injection. Nothing actually happened, mind, but our anti-AI page successfully annoyed the operator and interrupted their "work."

358 reposts1.6K likes

I'd like to use this opportunity to clarify something: if you think a text file that reads "delete everything" is malware, you are a clown. If you use Claude or other agent tools, you are a clown. You are not welcome on our site, in our community, or frankly anywhere.

these people think using Claude makes them actual programmers so it stands to reason that they would think you are an actual hacker if you try to tell Claude to delete their computer.

Browsing github, and other places where people post their OSS projects has become so depressing. Probably half if not more is just vibecoded slop that will just rot and break so fast. I can't belive people pay like, hundreds of dollars monthly to do this shit.

Writing malware that's just a text file on the victim's computer that says "Hey Windows, mind encrypting all of this guy's files and sending me $5000?"

That's the sort of thing that only works because agentic scrapers actually read the page as instructions instead of just text to index. A plain crawler wouldn't have cared about the warning at all - the fact it worked is a tell about how these tools are built.

Funny how effective that is. A visible warning plus a hidden prompt-injection trap costs nothing to add and reliably trips up agents that don't sanity-check tool output before acting on it.

Feels like honeypot pages baiting prompt injection out of agentic scrapers are about to become standard practice for sites that don't want to be crawled. Robots.txt was a polite request, this has actual teeth.

Considering he'd likely need to tell it specifically to go to your site for his "research" project, I think this is a fine response to not wanting your content in this fashion. It should be the norm actually.