learn-voicelisted

# Learn voice `writing-style.md` is the user's own description of how they write. Every drafting skill layers it on top of the default voice in `${CLAUDE_PLUGIN_ROOT}/docs/voice.md`. Until now it had no author but the user, typing it out by hand, so for most people it stays the blank template and every message comes out in the default voice. This skill fills that gap in one pass. The user supplies samples of their own writing, this skill derives the style rules from them, backs up whatever the file held, writes the derived file, and reports what landed and what it saw but kept out. No draft, no proposal to approve, no temporary file: the run ends with the real file written and the previous one recoverable. **Slash-only on purpose.** It reads the user's own prose and writes a file outside the project. Nothing about "make this sound like me" should make it fire on its own -- the user runs `/daikenja:learn-voice` when they mean to. `disable-model-invocation: true` is set for that reason. Throughout this skill, `writing-style.md` means whatever `profile.writing_style` resolves to -- a local file by default, or a Google Drive file. Steps 5 to 7 are where the two differ; everything else is the same work on the same prose. ## Where this sits next to `setup-user` `setup-user` Step 5 states a rule that this skill deliberately does not share: *"Already there. Leave it alone... Never inspect or overwrite user prose."* That rule is load-bearing for `setup-user` being safe to re-run