The time that I had a look at this package I also thought: why not an io.Writer?

But after that initial thought, I realized that the current API is very appropriate because:

checking for tty and applying wrappers to handle is something that must be done early in the program, usually in the main package, where you have access to raw os handles. The current API works fine for that use case. High level libraries (such as loggers) must instead work on abstractions such as io.Writer and it should not be their responsability to do dirty stuff like applying wrappers to OS handles.
the fact that you need to have access to the raw file descriptors to use the functions of go-isatty IS a feature as it blocks bad uses of go-isatty deeply in libraries. It forces users of the API to design the program in a way and that is a good thing because calls to go-isatty are costly because of syscalls.
adding a wrapper to allow easy access to io.Writer/io.Reader would just make it easy to use go-isatty incorrectly (for example on every call to a log method of a logger object) and that would be just bad for the whole Go ecosystem.
the proposed wrappers for io.Writer/io.Reader can easily be written outside of go-isatty. Making them available in the go-isatty module brings no benefits compared to making them available is a separate package, but brings the drawback of 3.

Allow to check a io.Writer instead of raw file descriptor #80

Description