Why do I want to do this? I need to know what the repolist was at the time of build. Instead of having that command in this test I opted for a simple file. Why write to a file and not just shell out to yum repolist -v directly? I like the idea of having the file on the filesystem of the image as well. Your needs may be different.

Build

# mount ./ as /hostfs and pass build arg for test #6
sudo docker build -v `pwd`/:/hostfs --build-arg TEST6=works -t jewzaam/test-docker-labels .

Results

# grab the test labels only.  requires jq
sudo docker inspect jewzaam/test-docker-labels | jq .[].Config.Labels | grep '"test'

Summary

You cannot set a label from the contents of an image at build time.
You can pass build args to set a label.
If you want to set a label from the image contents you have to inject it as a build arg.

Image contents into label example

# initial build, gets the content available for a second build.
sudo docker build -t jewzaam/test-docker-labels .
# file test.out is now written to root of image
sudo docker build --build-arg TEST6="`sudo docker run -it --rm jewzaam/test-docker-labels cat test.out`" -t jewzaam/test-docker-labels .
# verify
sudo docker inspect jewzaam/test-docker-labels | jq .[].Config.Labels.test6

Result should be content of the file written by the docker build:

"I was here\r"

ALTERNATIVE

You could write to a file inside the image to a volume mounted from the host then not have to do docker run to get it.