The only upside to a walled garden is that this type of stuff isn't supposed to get through. If it does, the store only servers to give users a false sense of validity.
I wish there at least seemed to be some degree of review or reasonable sandboxing here. The closest they come is disabling eval-style behavior in 'background' scripts, but there's nothing stopping you from running command & control scripts from a remote origin in a non-privileged context and then getting up to your evil mischief anyway. Or injecting malicious code directly into gmail tabs.