The system uses a hybrid blacklisting/whitelisting and AST-traversal approach to both remove dangerous builtins as well as disallow potentially exploitable python structures alltogether. Examples are while structures and attribute allocation. All advanced functionality is accessed through a set of "safe" methods on a holder object. You can extend this with your own safe methods in order to add more functionality befitting your game.
The system comes with a host of examples, a few scriptable objects and complete commands for adding code to objects. At this point it's not guaranteed that all systems are safe against meddling however - notably Attributes have no locks defined on them by default (although this system does properly check Attribute lock types should they exixt).
Please test and try to break - and report problems to the Issue tracker/forum as usual.
