voiceprintlisted

# voiceprint — build a writing voice from real samples, then write in it ## What this is Most attempts to make an AI "write like me" fail the same way: someone describes their tone ("friendly but professional") and the model generates a generic average of that description. The result sounds like everyone and no one. This skill works the opposite way. A person's voice already exists, fully formed, in text they have already written. The job is not to *invent* a voice from adjectives, it is to *read* real samples and extract the concrete, repeatable patterns that make their writing recognizably theirs: how they open, how they hedge, the punctuation they lean on, the words they would never use, the emoji that mean something to them. Those patterns get written into a self-contained **profile file**. After that, drafting in their voice is just applying the profile. The output is complete and reusable: a user points the skill at their own writing and walks away with a self-contained profile that makes every future draft sound like them. ## The one principle that makes or breaks a profile **Voice is extracted from real writing, never invented.** Everything in a profile must trace back to something the person actually wrote, or to something they explicitly confirmed. The moment you start guessing ("they're a designer, so probably casual"), the profile drifts toward generic and the whole product loses its value. Practically, this means: **collect real samples before writing a si