in-my-own-wordslisted

# in-my-own-words A voice is not a tone setting. It is a set of decisions the author makes over and over — how long a paragraph runs before it breaks, when a metaphor is allowed in, which word gets the italics, how the last line lands. This skill holds those decisions in a profile and applies them. Long-form only. Short-message register belongs to whatever always-on writing guidance the user already has; this skill takes over when structure and argument arc matter as much as word choice, and it should say so rather than fight for the same ground. ## Before anything 1. Read `config.yaml` in this skill directory. It names the author, the publication, the languages and where the profile lives. If it is still the unedited template, run **Bootstrap** below instead of guessing. 2. Work out the target language. For drafting and rewriting it is the language of the piece the user wants, taken from their request, not from the profile's primary language. For auditing there are two: the lexicon is chosen by the language of the **text under audit**, while the audit itself is written in the language the user is speaking to you in. 3. Load `references/core-voice.md` (language-agnostic) **and** the lexicon for the target language. Both, always. The core carries the structure; the lexicon carries the surface. Load `references/stated.md` too if it isn't empty. 4. **Precedence, when sources disagree: the corpus wins — if the corpus is theirs.** Check `sources/index.md` for provenance before