“Weasel voice” is a fake, possibly misleading name for a real problem encountered by lots of RPG writers.
Everyone has heard of active voice and passive voice in writing. (If you have not, Google it and come back.) In active voice, the subject acts upon the verb. (“She attacks.”) In passive voice, the verb acts upon the subject. (“They will be attacked by her.”) Passive voice produces longer sentences that sound superficially more sophisticated but are usually less interesting, so most people tell you to favour active voice. This is mostly right. There are times when passive voice is the better choice, though getting into why is beyond the scope of this little piece.
“Weasel voice” is not grammatically cut and dried, which is why it’s a fake, possibly misleading name. The special conditions of RPGs cause it to proliferate. Weasel voice is when, knowing that a given subject might do several things, you refuse to make definitive statements about even likely actions. The same thing applies to opinions, locations, and states. You fill everything with “weasel words:” qualifiers and turns of phrase that stop short of making assertions about what’s going to happen.
While you do have to cover the major interesting possibilities of play (and possible trouble spots), it is usually better to limit the set of options you present to just those interesting ones, using decisive language. Compare:
If Vath is backed into a corner, he will most likely use spells to summon allies, unless he believes a member of the party is weak-willed, in which case he may use his enchanted voice to gain their obedience. Vath will only attack with his dagger as a last resort.
If cornered, Vath casts summon allies to call his infernal crew, or uses his enchanted voice to attempt to ensnare a foolish-looking party member.
So, this revision has a few refinements. I turned passive voice to active voice in “cornered.” I got rid of the weaselly “most likely,” and “may use” bits. I got rid of “gain their obedience,” because that is already implicit in the power—Vath’s voice doesn’t set people on fire. I removed the statement about the dagger because game statistics should make it obvious that a silver-tongued evil wizard isn’t going to use his dagger first. I also made the tactics description more specific, naming a spell instead of alluding to it.
The revision is shorter, but that’s not why it’s better. It carries more useful information in a more readable package.
The critical trick here is to rely on a sense of what is self-evident. You do not have to repeat what a spell does if the description is somewhere else. You don’t have to say someone who sucks at using a dagger is going to favour a different attack. Most of all, you can assume these are a few notable possibilities among many. You almost never need to introduce qualifiers to let the reader know other even less likely options exist, because that’s always true in every game.
When writing for beginners or dealing with a bunch of edge cases, you will have to get a bit weaselly, but most of the time people know that Facepunching +10 means somebody’s mostly going to punch people in the face, but might have a chat with you instead, or might do whatever else the player or GM imagines, within the bounds of the game.