If I understood your question correctly, I was able to achieve this with littler , which I use instead of Rscript for script tasks that revolve around R. I run CentOS 7 and, based on the code in your question, it looks like you're on a Unix-like machine, so installing littler should not be a problem. For minimal reproducibility, I used the standard brilliant application and the brilliant Rmarkdown templates provided by RStudio, saving them as testapp (project / application directory name) and testRMD.rmd respectively. Then I have the following scripts:
testapp.r
#!/usr/bin/env r shiny::runApp( "~/tmp/delete/testapp", port = 7088, launch.browser = TRUE, host = "127.0.0.1")
testRMD.r
#!/usr/bin/env r rmarkdown::run( file = "testRMD.rmd", dir = "~/tmp/delete", shiny_args = list( port = 7088, launch.browser = TRUE, host = "127.0.0.1"))
Set permissions for these files so that they can be executed -
[ nathan@nrussell R]$ chmod +x testapp.r testRMD.r
( chmod +u ... should be enough, but independently ...), and you should be configured to start from your terminal, etc.
[ nathan@nrussell R]$ ./testapp.r Loading required package: shiny Listening on http:

[ nathan@nrussell R]$ ./testRMD.r Loading required package: shiny Listening on http:

There is additional command line output for the Rmd file that I skipped, but I am sure that this can be easily suppressed if necessary. In any case, it seems to be working fine - both the brilliant application and the Rmarkdown application are interactive, as when starting from RStudio, but if you have anything else, please clarify.