For years, “third-party.com” has been one of those made-up web addresses that fill documentation and code samples, like example.com. Someone registered it, and it now serves a ClickFix trap to Windows visitors, The Hacker News reported on 24 September, citing research by Manifold Security. Two similar placeholder addresses show Mac visitors scareware and an investment scam.
What happened
Unlike ‘example[.]com,’ third-party[.]com is not IANA-reserved. Anyone could register it, and someone did. Every doc, test, and skill that hard-coded it now points readers at attacker infrastructure.
Ax Sharma, head of research at Manifold Security, quoted by The Hacker News
According to Manifold, the domain has served its lure since at least June 2026. Windows visitors see a fake Cloudflare check that quietly puts a command on their clipboard and tells them to paste and run it; the command fetches and runs a remote PowerShell payload. Mac visitors get an error saying macOS isn’t supported. A GitHub search finds the domain in more than 1,700 public repositories, including AI agent skills and documentation for MCP servers (the plug-ins that give AI assistants tools) that cite it as an example endpoint. It is now flagged as unsafe on VirusTotal and Google Safe Browsing.
Manifold then found 13 more placeholder-style domains that aren’t reserved. Two of them, yoursite.com and your-domain.com, show macOS visitors scams while other visitors see an ordinary parking page. On a Mac browser, security researcher Cody Nash said, your-domain.com displayed a fake “MacOS Security Center” claiming four viruses and selling a counterfeit McAfee renewal, and yoursite.com showed a counterfeit ZDF news article advertising an investment scheme. Those two appear in hundreds of thousands of GitHub files and hundreds of agent skills.
Why it matters, in plain words
People skim past example links, and so do the tools that check code. As Manifold points out, a file scan can’t see what a website decides to send: the page looked harmless from their analysis machine and only turned hostile for the right kind of visitor. Now that AI agents read documentation and follow links on our behalf, a “harmless” placeholder can become an instruction source nobody meant to trust.
What the Mac version of this trick usually looks like
The Mac visitors in this case got scareware rather than a paste-this-command trap, but the command trick has a Mac form too, and it is worth recognising. A page tells you to open Terminal, then paste one line to “fix”, “verify” or “install” something. The line is usually unreadable on purpose: a long string of letters and digits that gets decoded, or a short command that downloads a script from a web address and runs it straight away.
The reason it works is simple: when you type or paste a command yourself, macOS treats it as your decision. Gatekeeper checks apps you open, not every line you run in Terminal. No legitimate website, support page or security check needs you to paste a command into Terminal to view it. If one asks, that request is the warning sign.
What to do
- Never paste a command a web page puts on your clipboard, whatever it claims to fix. That is the whole ClickFix trick, on Windows and on Mac.
- A page that says your Mac has viruses and offers to sell you protection is a scam. Close the tab.
- If you write documentation, tests or agent skills, use the reserved names example.com, example.org or example.net, and treat plausible-sounding addresses like yourcompany.com or your-api.com as something anyone can register.
- If you run AI agents or MCP servers, audit the links in the skills and docs they load.
Where FireAI fits, and where it doesn’t
- FireAI shows every connection each app makes, including AI assistants and the local tool servers they launch, on the world map. An agent reaching a domain you’ve never heard of is visible, and you can see which app did it.
- You can block a domain for one app or for every app, so a known-bad placeholder stays unreachable from your Mac.
- New apps and command-line tools ask before their first connection; in Paranoid mode even Apple’s own tools must be approved, the ones a pasted command would use to download a payload.
- Clipboard Armor flags an app’s first connection that lands seconds after your clipboard changed, the timing a clipboard-poisoning page produces.
What FireAI doesn’t do: it doesn’t read web pages or judge what a site shows you, so it won’t spot a fake security alert inside your browser. That part is on you, and now you know the tell. Download FireAI and try it free for 17 days. FireAI is made by HisnLabs.