Woah, looks like this was made by Steven Lavelle (http://www.increpare.com/) too, who has made a staggering number of bizarre and fascinating indie games over the years. My personal favorite is Mirror Stage, where you navigate patterns in a kaleidoscope:
PuzzleScript is an impressive project. The editor is pretty easy to use (the smart syntax highlighting is very helpful). The rule system was general enough to allow me to implement most of the concepts from the original Pegs very easily (with the exception of cross piece selecting), and add new features like animation and sound effects easily too. It was tough to compress some of the 8x8 sprites into 5x5, but I managed it thanks to the ability to use custom colors in sprites. What I didn’t like was the hard-coded limit of 6 layers, plus the occasional severe visual glitches in the running game that required me to refresh the page. The documentation is very good for a one-man project, but has small errors in places and lacks cross-references that force me to look through many pages to find the functionality I was looking for.
I was very interested when I saw this post, because I was already re-remaking Pegs for the browser (https://github.com/roryokane/pegs-js). I have a lot of experience with implementing that game, so it was interesting to see the rules phrased in a different language, and to come up with new colored sprites, sound effects, and animations that reflect what I imagine is really happening when you play Pegs.
I reckon you could go a bit beyond simple tile-based puzzles with that beautiful logic engine. A boulderdash-type game seems feasible, but what about Dwarf Fortress?
Someone found 1. a common denominator of a family of Sokoban-like games, encoded them in the form of 2. a simple pattern-matching language, and implemented 3. a game engine on top of it.
So read this if you are interested in either 1 or 2 or 3.
I have some puzzle game ideas that I've tried to build several times, but I always get frustrated with how tricky it is to express the rules in code. This seems like such a clever way to do it. Fun!
I'm not sure what you're getting at with "useful," but it looks like it's going to be incredibly fun to play with, and I wish it were available when I was a kid messing around with HyperCard stacks.
I spent an hour trying to implement a 2d Turing Machine.
http://www.puzzlescript.net/play.html?p=6863139
So far, it is unsatisfactory because I do not know how to emulate an infinite playfield, and I cannot figure out how to put a delay between applications of the rules (so you cannot see it running).
That said, with a little tinkering, a Turing machine is possible.
http://www.increpare.com/2009/03/mirror-stage-done/