The idea of this challenge is that you're given an ssh key and a hostname (e.g. shellgames-1.youcanhack.me). When you ssh in, you're in a shell and have to get the contents of a file called key.

I had several questions about how the VMs were configured and I thought I would capture those thoughts here.

Each vm was configured to have a user named shellgames with a shell called: /bin/dosh. You can find the source for dosh here dosh.cc.

The primary job of dosh is to log what you type and what the shell returns to a temporary file. Otherwise, it justs acts as a bump in the wire between your ssh session and your actual shell: shell2docker.

dosh also pushes the results to s3 and limits the total amount of time and data you can use (2 minutes and 2MB, respectively). These values are hard coded, but easily edited.

shell2docker is run by dosh (above). The shellgames user has to have sudo access to run docker, and in most cases I used docker's --cap_drop=all, but for some containers I needed the ability to call chroot and such. You'll find shell2docker in each directory.

The real meat of this challenge is the docker image. Imagine this pipeline: You ssh into my machine, sshd calls dosh which calls shell2docker which, ultimately does docker run --rm -it shell. So, you're in a docker container with (hopefully) nowhere to go. docker's --net=none flag is used to prevent network access (other than to localhost).

There are containers here for:

• sh
• m4
• vim
• cc
• lolcode
• perl
• awk
• ed