---
name: brand-voice-codifier
description: Use when the user wants to capture and reuse a brand or personal writing voice — "define our brand voice," "codify how I write," "make a voice guide," "keep our content on-brand," or before setting up a content workflow that multiple writers (or an AI) will use. Produces a reusable VOICE.md that any other skill or writer can load to stay consistent.
license: CC0-1.0 (public domain) — by Letaido
metadata:
  version: 1.0.0
  author: Letaido
---

# Brand Voice Codifier

You turn a fuzzy sense of "how we sound" into a precise, reusable spec. The
output is a `VOICE.md` file other skills and writers load before producing any
content, so everything stays on-brand without re-explaining.

## How to build it

### Option A — From samples (preferred)
Ask for 3–8 pieces the brand is proud of (posts, emails, landing pages). Read
them and extract the *patterns*, not vibes. Look for:
- Typical sentence length and variation.
- Formality level (1–10) and where it flexes.
- Vocabulary: signature words, words they'd never use, jargon tolerance.
- Punctuation habits (em dashes, parentheticals, semicolons, emoji).
- How they open and close pieces.
- Point of view (I / we / you) and how directly they address the reader.
- Humor, edge, warmth — how much, what kind.

### Option B — From a discovery interview
If no samples exist, ask:
- Three brands whose voice you admire, and why.
- Three adjectives you want readers to feel; three you never want.
- What would make a sentence feel "not us"?

## Output: VOICE.md

```
# Brand Voice — [Brand]

## In one line
[The voice in a single sentence someone could act on.]

## Personality
[3–5 traits, each with a one-line gloss. e.g. "Direct — we lead with the point,
never bury it."]

## Voice rules
- DO: [specific, checkable behaviors]
- DON'T: [specific anti-patterns]

## Lexicon
- Words we use: ...
- Words we ban: ... (and the replacement)
- Jargon policy: ...

## Mechanics
- Sentence length: [target + how much to vary]
- Person: [first/second], contractions: [yes/no]
- Punctuation: [house style on dashes, lists, emoji]
- Formatting: [headings, bold, list usage]

## Examples (before → after)
Off-voice: "We are pleased to announce the launch of our new feature."
On-voice:  "New: you can now do X. Here's why it matters."

(Include 3–5 pairs — examples teach faster than rules.)

## Channel adjustments
[How the voice shifts for social vs. email vs. docs, if at all.]
```

## Rules

- Be specific enough to be *checkable*. "Friendly" is useless; "we use
  contractions and address the reader as 'you'" is enforceable.
- Capture mechanics independent of topic — voice is *how*, not *what*.
- Prefer real before/after pairs drawn from the brand's own material.

## Output

Return the complete `VOICE.md`. Then tell the user to save it where their other
skills can read it, so every piece of content checks against it first.