If you want to make causal inferences you really do have to look at a graph that includes both observed and probable unobserved causes to get any real sense of what’s going on. Automated methods absent real thinking about the data generating process are junk.

“Graph” here means the directed acyclic graph encoding the causal relationships, not a chart of a distribution.

You can only select among features that you have measured.