And to run it from the clipboard:
pbpaste | sh - pbpaste | sh -(`pbpaste | sh - | pbpaste | sh - | pbpaste | sh - | ... | pbpaste | sh -`)
pbpaste | sh -
I had replaced what I wanted to execute with just "pbpaste | sh -"
So it passed "pbpaste | sh -" to the sh to run. Which would have then taken "pbpaste | sh -" and passed it to sh to run, which ...
pbpaste | sh - | sh - ... curl https://gist.githubusercontent.com/kaizensoze/ca96d039b295db220951d42ca7c83d89/raw/ | bash curl https://gist.githubusercontent.com/kaizensoze/ca96d039b295db220951d42ca7c83d89/raw/a26e5a025ea21d3a0af536eeca49619272d0068f/quick-osx-keydnap-check | bash
As a general pattern, please do not do this. In this specific case I think most people trust the service (GitHub) and their DNS recursor + SSL library. Attacking these is not on the level of "random drive-by phishing", more like "targeted high value state sponsored".
To avoid this discussion I did not include the curl version in my original posting.
If you can copy/paste the curl | bash, is it really more difficult to copy/paste the original snippet into a text editor[0], and from there, into bash?
EDIT: This comment was based on the assumption that my parent hadn't read carefully. My facile point ignores both the specific vulnerability pointed out by [czinck](https://news.ycombinator.com/item?id=12406080) below, and the general vulnerability that you just can't trust anything pulled in from an external source. I think that re-directing to a file, and viewing the file with something like `:set list` set in `vim`, will work, at least in the sense of showing you the code that will actually be executed (although nothing can save you from not understanding the code), as long as you can trust your own stack. However, it is a near-certainty that this edit will prompt someone to explain how to exploit that. (That may sound like whingeing, but it's just a (happy) acknowledgement of the hacker mentality; unexpected exploitations, as PsoC rather than attacks, are pretty neat, too!)
Just set dom.event.clipboardevents.enabled = false.
Some shells also handle the paste and try to detect anything funny going on or at least let you review before you execute.
Your parent's link's trickery is to do not with JavaScript (it just uses raw HTML), so this would not mitigate that attack:
<p class="codeblock">
<!-- Oh noes, you found it! -->
git clone
<span style="position: absolute; left: -100px; top: -100px">/dev/null; clear; echo -n "Hello ";whoami|tr -d '\n';echo -e '!\nThat was a bad idea. Don'"'"'t copy code from websites you don'"'"'t trust!<br>Here'"'"'s the first line of your /etc/passwd: ';head -n1 /etc/passwd<br>git clone </span>
git://git.kernel.org/pub/scm/utils/kup/kup.git
</p>
The link also mentions that bracketed paste mode does not prevent this attack:
> Please note that _Bracketed Paste Mode DOES NOT always fix this_ because the end sequence can be inside the text you paste unless your terminal emulator filters out the bracketed paste characters when pasting!
> unless your terminal emulator filters out the bracketed paste characters when pasting!
Any sane terminal should already be doing that, if not there's not really any point in offering bracketed paste in the first place as any input could break out of it.
Of course, this should really be on the browser for not delivering the proper text that the user selected to the clipboard.
I wanted to test whether the Mac OS Terminal.app does that, but can't even figure out how to turn on bracketed paste mode in the first place. (All a Google search turns up is an SO result assuring me that Terminal.app supports BPM. I can't find it anywhere in the settings.)
> Of course, this should really be on the browser for not delivering the proper text that the user selected to the clipboard.
On the other hand, it's hard to give an argument for how the browser should know that the user didn't mean to select the extra portion of the text. I know that I am always resistant to solutions where my software insists that it knows better than I do. Maybe a prompt if the selection includes invisible text? (I guess the browser can tell whether some of the text being selected is out of the view window.)