I have found a solution for Mastodon. If you use a nsfw hashtag, the image is also hidden on Mastodon.

@Montag I think it'd be good if when publishing comments to masto that it adds a #nsfw tag, if that means images are hidden...

I think that images should also remain hidden in a post with an abstract.

What I'd really love to see is for our markup to look more like this:

[abstract txt="This is the 1st CW message"]
This is the text that is protected
[img ...]This image is not shown unless the user chooses[/img]
[/abstract]

This text is not protected

[abstract txt="This is the 2nd CW message"]

This is more text that is protected

[img ...]This image is not shown unless the user chooses[/img]
[/abstract]

This would allow one to compose a post with multiple CW sections, each of which could be shown / hidden independently, including any images inside the [abstract] tags.

I realise not all instances will understand this markup, but presumably it could be translated for each instance, as the post is generated for sending, so:

* A masto instance might receive:

[abstract]This is the 1st CW message[/abstract]
#nsfw
This is the text that is protected
[img ...]This image is not shown unless the user chooses[/img]

This text is not protected

This is more text that is protected

[img ...]This image is not shown unless the user chooses[/img]

And thus only show only the first abstract and nothing else unless the user chooses to.

* A friendica node would receive it unmodified, and allow the user to selectively show / hide each block protected by its [abstract] tag.

etc.