brand-voice-guidelisted

# Brand Voice Guide Extract how a brand actually sounds from what it has actually written, and turn it into a directory that a copywriter, a new hire, or a Claude agent can produce on-brand content from. This skill is opinionated about one thing: **a voice guide is extracted from real writing, not invented from adjectives.** "We're bold, human, and authentic" describes nothing. A verbatim line the brand wrote, and a rule named from it, describes everything. The output is a **directory of reference files** (not a single essay), so it can be consulted modularly at the moment of writing — the voice-core rules always, the relevant channel file for the piece at hand. It mirrors the structure a production content system consults before generating a word. ## What this builds (and what it doesn't) - **It builds** the voice directory: `voice-core.md` (the character), `writing-standard.md` (the concrete rules — a *Do* half and a severity-tagged *Avoid* half), per-channel files, `real-examples.md`, plus light `about-company.md` / `icp.md` context. - **It does not check drafts.** Governing new content against the guide (AI tells, template language, brand-rule violations) is the job of a separate `content-quality-check` skill. This skill produces the standard that skill enforces — the *Avoid* section of `writing-standard.md` is the part it reads. ## Inputs (harvest first, ask only for the gaps) The skill is harvest-first: it extracts as much as possible from existing writing