This card is an interesting one because how good it is depends heavily on so many factors it can be hard to evaluate. That said, I ran some initial statistics based on cards such as Olive McBride and Dark Prophecy.
I think aramhorror gave a solid analysis of the capabilities of Song of the Dead if you're not able to get those . It's not worth it. However, the is a part of the card and the ability to control the chaos bag has evolved. I wanted to flesh out the analysis of the probability of getting that in different scenarios.
Each campaign has slight variations in how many skulls are included, and these have a great effect on the default chance you'll draw a without any chaos manipulation. You'll usually find ~18% chance. If you're Jim Culver, you're looking at an extra 6% chance. Basically, 1 out of 5 times, you'll just do 1 damage.
But you are no bystander! You are a mystic will the power and skills to change probability to suit your whims!
The greatest ally to this card is obviously Olive McBride. She modifies your chances of a skull from ~20% to ~50%. This means, on average, you're dealing an extra 1 damage when you attack. This puts it on par with Shrivelling.
But that's not all. For those moments where you absolutely HAVE to get that token, you can use Dark Prophecy in conjunction with Olive McBride. You end up drawing 7 tokens and you have anywhere between 64%-85% chance of getting a skull result. By sealing some non-skull results, you can boost your probability for success slightly higher.
Then there's sealing with cards like The Chthonian Stone, Protective Incantation, and Seal of the Seventh Sign. Sealing provides a slight boost to success rates depending on what tokens are in your bag. By and large, sealing any non-skull token gets you about 1% improvement to base probability, but it compounds up to about a 5% improvement with both token drawing options active. Note that the The Chthonian Stone will have no effect for the first couple scenarios of Carcosa.
You can further affect the chaos bag by purchasing a Grotesque Statue which can further improve your odds.
The very best chances you'll ever get is up to a ~95% success rate with Jim Culver if you seal 3 non-skull tokens and the .
But at the end of the day, we have to look at the cost. Each of these cards costs resources, card space, and actions to play. To full support this card, it'd take 14 resources, 6 cards, and 6 actions to play those cards. That doesn't include the cost of Song of the Dead or the upkeep for Protective Incantation.
Ultimately, your greatest gains are going to be from cards that let you draw more than one token. Sealing tokens helps, but it's often too costly to be worth it in the long run.
The point is, this card isn't bad, but it's not exceptional either. With Olive McBride, it's on par with Shrivelling. It averages to the same damage with slightly less but no downside for drawing a bad token. With Dark Prophecy, it surpasses Shrivelling, but only for the two tests you use with it.
TLDR: In a deck that's built for it, this is slightly better than Shrivelling. Otherwise, go for Shrivelling and it's successors because they upgrade into each other and are more reliable.