The rule that the config should hold the name and never the value is the clearest version of this advice I have seen. One git add is all it takes for a pasted key to travel everywhere the repo goes, and a gitignore entry only hides that problem instead of fixing it. The scope guidance is practical too: if a deployed agent needs the key, put it in the team scope, since a team deployment never reads the user scope. I also liked that rotation is just editing the variable and restarting, with no redeploy. The honest note that scoping does not isolate a key from the run that uses it is important. A key a run can resolve is a key the run can leak.