By default, Windows batch files are run in the parent shell's context (which is unusual for Unix users, where an explicit source is needed, but was the only possibility in MS-DOS). This means directory changes and environment variables affect the original interactive shell too.

Put setlocal at the top of your script to make it run in its own context – you can safely use cd inside the script then.

As grawity previously mentioned, pushd \windows && (foobar & popd) would work better than pushd \windows & foobar & popd because the latter may fail if there is no such directory.

Also, using setlocal and endlocal allows you to have multiple local environments, so for example you could have:

setlocal

cd dir

command

endlocal

Now you would be back in your original directory.

You can use cd - to go back to the previous working directory. And use ; instead of &&, then the subsequent commands won't be dependent on the success of previous commands.

$ pwd
/etc
$ cd /var ; pwd ; cd -
/var
$ pwd
/etc

I applaud grawity’s suggestion to put setlocal at the beginning of your batch script, but I would add the fact that you can have multiple, nested, setlocal / endlocal blocks, so a more relevant answer to the question might be

@echo off
setlocal
cddir1
  ...
setlocal
cddir2
command
endlocal
:: Now I am back indir1
  ...

And, of course, if you want the command to be executed only if the cd to dir2 is successful, say cddir2&&command.

Note that the setlocal / endlocal block creates a localized environment, so any variables that you set or change in such a block will revert to its previous value after the endlocal.

You can save the current directory into a variable. Change and change back depending on the return value of the command. BTW, %CD% returns your current DIR.