I would love to hear what restrictions you think you can place on a Turing complete brower extension (a model that already has tons of permissions that apparently aren't enough) would let you know that a sufficiently large subset are sufficiently safe to allow $5/app to be enough to subsidize screening the rest.
In this case, this is a browser extension, and the goal is to edit the page: there are tons of such extensions; while a couple of these had obvious "talks to the wrong server" problems, various of them were correctly only editing pages on Facebook, but were doing so maliciously.
The price point we are talking about here is simply so low: $5 is only going to purchase 40 minutes of a minimum wage employee... we aren't even talking a junior supervisor at a fast food restaurant, we are talking about the entry-level "try to get the orders in the computer right" position.
At these prices, even just reading the description and figuring out "oh, this should only be able to edit pages on this one website" is already going to be expensive. Figuring out "should only be using DOM to remove nodes and not add script elements" is impossible.