These cards should also be hidden from the atom feed.
A label could be anything, most likely a size-limited field:
"Spoilers about the hobbit" "Spoilers" "Sexual content" ....
For example, If I have a stream where I review movies as opposed of 1 stream per movie, I could mark a specific card as containing spoilers about a specific movie. In this case, a reader who has already seen that movie could read the card.
Another way to do it, and perhaps a more generic one, would be to have a 'sensible content' label that could be shown as a lightbox - not a designer, forgive my lack of good terms - on top of the card with a custom message.
This label could be use to hide cards that the writer considers contain sensible content that could affect others in some way. Opening them or not is up to the reader.
@jtauber started a new stream about how 'spoilers' warnings could be implemented in thoughtstream. One way to do it, the one he's proposing, is to have a flag on streams or even thoughts themselves that would mark them as spoilers.
In this case, readers would be able to choose whether they want to open it or not. It would be also possible to hide the thought from the front-page.