There are multiple techniques I've seen recommended for handling prophecies (precognition etc.) in RPGs. You have the option of either sticking to one of them, or combining them.
Path of No Resistance
The prophecy describes the way things would become if nobody knew the prophecy. In other words, the very act of viewing the future is likely to change it. Of course, some events may or may not be easily changeable. It's quite possible for an aversion of a prophecy to be contingent on a very difficult series of actions that have been set in motion by the discovery of the prophecy, and whose success or failure is still in flux.
This is probably the best technique to use if you want to be able to provide prophecies at a moment's notice, since you probably have a general idea of what future events are likely. Take the general idea, pick a variant thereof that is more specific and can be averted, and present that as the prophecy.
A Direct Twist
Something is predicted (usually in a verbal form, occasionally in a poetic one too), and the predicted things are set in stone. However, what exactly does the prediction mean is ambiguous enough that it's possible to achieve it in multiple ways. E.g. 'You will conquer all of the observable universe a day before you die' could mean either conquering everything, or being locked in a minuscule pocket dimension; or it could mean that the world will be so desperate to defeat you that the other powers will ward off a small part of the universe and eject it into a pocket dimension (like the Orions did with the Antarans in Master of Orion). There's a lot of space for variation.
Beware: direct word-twists can be disappointing, and feel like 'gotchas' if the players don't like word games. It also can result in reducing agency by reducing ability to make informed choices (an uninformed choice can often be as bad as no a coinflip1). Also beware that this technique doesn't play nice with sudden requests to provide a prophecy unless you knew what will be asked about.
But if your players are aboard with such a style, go for it.
Twist of Context
The facts shown (and they're usually shown, not described) are set in stone, but there's very little information on what surrounds them, which may make them not that bad, or not that good, or whatever. I've seen a similar principle offered for resolving certain time travel paradoxes (e.g. in GURPS Infinite Worlds).
This kind of twist is probably more satisfying to players who like to have a more direct, hands-on influence on what comes to pass, and it feels like less of a 'gotcha'. This works OK with sudden requests of a vision if you know for sure and in advance what set-piece scenes you'll want to present to players.
Prepare for the Worst
The prophecy will come to pass as predicted. However, knowing this will still give the PCs an opportunity to prepare for it and mitigate the consequences of it. E.g. you know about the incoming asteroid impact, and that it cannot be averted, but at least now can start building an ark-ship.
It's pretty easy to provide this one on demand if you know what major cataclysms are planned in the campaign (e.g. you plan a dino-killing asteroid for the third act of the chronicle).
The prophecy sets in motion events which make the prophecy possible. This variant can be either set in stone, or the prophecy can be about a potential future. Usually this works best with positive prophecies (e.g. 'You are destined for greatness, Po!').
Keep in mind that these are often tricky to set up if they depend on PC actions, and can lead to an urge to railroad if they don't work out at first.
Overall, this type of prophecies is extremely hard to implement without a lot of planning and without knowing when and how it will be presented to the PCs and their players.
A setting can either have only one of the above prophecy types, or it can combine them freely. Perhaps there are multiple types of prophets, or perhaps each school of prophecy can only access prophecies of one of the types.
1 Related: For the same reason, vagueness (whether verbal, symbolic-visual, or insufficiently informative but concrete visual) can make a prophecy feel pointless: you can't use your knowledge of the future if you don't have actual knowledge, just a string of words.
Sarah Connor Chronicles has an example of just how frustrating a vague omen can be: Sarah is convinced that a triangular pattern (three points, three lights etc. in a vaguely-triangular configuration, which is most configurations three dots can be in) has a deep meaning, and starts seeking said deep meanings in anything that matches the pattern. Most or all of it is a wild goose chase and she's beginning to doubt her ability to find meaning in this hunt. This is what players are likely to feel like if presented with a pattern that can be applied to anything or nothing. (Also, humans are notorious for seeing patterns where none exist, leading to many false positives in all the wrong places.)