Without the right car to test on, programming the manufacturer-specific messages is essentially coding blindfolded. I've got some information from Scangauge which I have roughly reverse-engineered, but the true test will have to wait until I have access to the car. Or a Scangauge, which may be able to confirm the (in)accuracy of my reverse engineering / forum searching.
Frank's MCP2515 library is easy to work with, so I'm pretty confident that what I program will make it onto the CAN bus, but is it the right thing? What will the response look like? The software will probably have to go on hold until I can test on the right vehicle.
Next step: assemble everything in a nice neat package. Might add a "buzzer silence" button too.