For instance:

 limit := <- api.GetRateLimitInfo()
 twuser := <- api.GetUser("hoisie")

Just curious why you chose this model. I'm having a hard time coming up with a use case where it's advantageous to delay receiving the data.

/Users/mike/apps/bin/8g -o go.8 api.go status.go user.go hacks.go search.go util.go rate_limit.go http_auth.go
http_auth.go:68: not enough arguments to CALL
make: *** [go.8] Error 1

You need to change the line to this:

resp, err = http.ReadResponse(reader, req.Method)

The following code prints 0, when it should print the seconds when the status was posted:

 package main

 import "twitter"

 func main() {

    api := twitter.NewApi()
     api.SetCredentials( "user", "pass")
     twuser := <- api.GetUser("user")
     status := twuser.GetStatus()
     println( status.GetCreatedAtInSeconds()  )
 }

In my twitter crawler I've been noticing that some messages contain xml char data tags like &lt, &gt, ൟ e

Go has a parser than can convert these to normal strings. However, it's part of the XML unmarshal function.

Is there a strict reason why you're using json as the encoding mechanism? It seems that if you make xml requests, the data would be converted more efficiently to a Go string.

Go-Twitter doesn't compile on the current release branch of Go
You need to run gofmt -oldparser on the source tree