Question: What should I do if I forgot which sample indices I used during library construction?
Answer: First, use
mkfastq to demultiplex with arbitrarily chosen 10x sample index set names just to get some output.
This will most likely not generate usable FASTQ files (unless by chance you happened to choose a 10x sample index set name that you actually used). However, it should produce a directory in the output called Stats, which should have a file called DemuxSummaryF1L1.txt in it (one such file for each lane).
In this file, there is a table at the top that shows the percentage of clusters in each tile assigned to each sample in the sample sheet (including Undetermined, which is likely to be very high in this case). Below that, there should be either a list of the frequencies of occurrence of each index that was sequenced, or a similar list of just the ones not able to be assigned to a sample (which are close to the same thing if nothing in the sample sheet actually matched). The specifics of the file depend on the version of software on the sequencer, or the sequencer.
From the list of sequences that were sequenced, you can then use the following spreadsheets to determine which sample index sets were actually used:
Cell Ranger for 3' GEX, 5' GEX, V(D)J
Long Ranger or Supernova for Genome
However, this doesn't tell you which sample is which, so once you populate the sample sheet with the correct 10x sample index set names and re-demultiplex to generate productive FASTQ files, you will still need a strategy for identifying which sample is which (depending on the situation, this may not always be possible).