You should be able to just pick the version that matches Spark and Scala from maven.

Here is a simple way to get the cluster Spark version

The question is about an init script though

We can infer the cluster DBR version using the env $DATABRICKS_RUNTIME_VERSION. (For the exact spark/scala version mapping, you can refer to the specific DBR release notes)

Sample usage inside a init script,

DBR_10_4_VERSION="10.4"
if [[ "$DATABRICKS_RUNTIME_VERSION" == "$DBR_10_4_VERSION"* ]]; then
  echo "running 10.4 specific commands"
else
  echo "Skipping 10.4 specific commands"
fi