My suggestion would be to add a test right at the top of the function and then either exit and ask people to load the settings first [then also need to update documentation] or invoke RmbDefaultSettings() and continue with a warning.
The problem with the latter is that people should ideally be doing the pre-screening using the settings they will then use for the RMassBank workflow.
Alternative: settings file is an input. Thoughts?

Absolutely the former. We must not assume default settings are the right settings.

@meowcat do you know offhand how to test whether settings are loaded in RMB?
Can't find it in the documentation

https://github.com/MassBank/RMassBank/blob/master/R/settings_example.R#L1

yes, though that function is of course not exported, so either ::: or directly is.null(getOption("RMassBank")).

Without looking at your specific code, I suggest this pattern we use in RMassBank:

findMsMsHR.direct <- function(msRaw, cpdID, mode = "pH", confirmMode = 0, useRtLimit = TRUE, 
			ppmFine = getOption("RMassBank")$findMsMsRawSettings$ppmFine,
			mzCoarse = getOption("RMassBank")$findMsMsRawSettings$mzCoarse,
			fillPrecursorScan = getOption("RMassBank")$findMsMsRawSettings$fillPrecursorScan,
			rtMargin = getOption("RMassBank")$rtMargin,
			deprofile = getOption("RMassBank")$deprofile,
			headerCache = NULL) { do.stuff() }

In this way, it is never opaque that there are parameters being accessed in a specific function call. Doesn't need to be as granular as this, settings = getOption("RMassBank") would be simpler.