Ah, sounds like you basically want the opposite of https://github.com/db0/threativore/tree/main/threativore - this bot finds things based on regex matching, and then "reports" it - i.e. report as spam to the moderators/admins. You instead would want it to take a different action - I guess retooting it to you (or I suppose it could also email you a link to the toot).
I think, and keep in mind I haven't ested this, but for a dev, it seems like this is still pretty easy to do by modifying threativore - just change this line here, https://github.com/db0/threativore/blob/main/threativore/threativore.py#L339 from the moderation-reporting call to the "notify me about interesting posts" call.
Thinking it over, it would be very easy to do this actually.
Not saying it's a good idea, just that it's technically possible - and made easier by the technical details of federation.
A modified, rogue instance could scan each post/comment/toot as it federates in against a list of keywords, and use a bot account to report the ones that fail the scan back to the original instance.
Worse, this rogue instance might not mark the account as a bot.
Interesting. So fediverse the term was only coined (2012) after the first instance by Mozilla was shut down (and it seems that there were many protocol changes in between).
And also ActivityPub was a few years in the future when the term was coined (2014 being the first mention).