So I did some playtesting on this, and not only does the mechanic seem to work well, but -- somewhat to my surprise -- the card I suggested, which I named Valkyrie, seems to work, too. Appropriately, sometimes it's the thing to do, and sometimes it isn't. If you're on a board where you can set up a deck that draws itself every turn (i.e., Hunting Party, Laboratory) without incurring curse tokens, it's probably a huge mistake to use it -- unless you can get it running a whole lot faster. If you can capture a sufficient lead, you're probably all set, because the +3 Cards keep the deck more resilient as it greens than, say, a Laboratory engine. But by and large, Valkyrie is a mistake in such kingdoms, as the curses just prove too much of a deficit to overcome.
However, it works very well in kingdoms with limited action chain alternatives. I've tried spamming them, and I've tried using small numbers strategically. Both seem to be viable strategies sometimes. Spamming seems to carry considerably more risk, as you'd expect. I can't tell yet, but it seems like strategic uses of the card performs a little better than spamming them, but I suppose it depends on the situation.
I also tried another such card, this one not yet tested quite as much:
Chimera
$6
+2 Cards, +2 Actions, +$2. When you play this, gain a curse token.
Similarly, this works best in kingdoms without an alternate source of extra actions. It's also even more important than with Valkyrie to make the decision about using them sparingly or spammingly. Without another source of +Buy, extra money is wasted; with +Buy in the mix, it's a powerhouse. But it still probably loses to sufficiently powerful curse-free-token alternatives, if any. Even if not, there are still interesting decisions to make about how many to buy, what to flesh the rest of your deck out with, and when to actually play the ones you draw in your hand. (And if you don't play them, as play2draw points out, there's a cost to that, too.)
The interesting thing to me about these cards is you really do have to think before playing them. Whereas playing Markets and Laboratories are no-brainers -- it's rare that automatically and unthinkingly playing those cards is a mistake (or a very big one if so) -- but with these cards, you really have to think about whether you have any alternatives to playing them, or if earning what they give you is going to mean buying better engine/VP cards or just supplying extra cash beyond what you can use.
So, to sum up, I really like the flavor this idea adds to the game.