Official skill · Marketing

Brand voice guide and rewrite

v1.0.2 · Updated 4 November 2026

Extracts a voice guide from 3 to 5 sample texts (traits with do and don't pairs, a word list, mechanics) and rewrites a draft in it. Use it when copy from different writers or AI drafts sounds like different companies.

Works in Claude, ChatGPT, and coding agents. On Team and Business, the official skill in your pod updates itself when we improve it.

When to use it

• Copy from different writers or AI drafts sounds like different companies
• You have a draft and want it to "sound like us"

What you give it

• 3 to 5 sample texts the team likes
• A text that sounds wrong (optional)
• A draft to rewrite (optional)

What you get

A one-page voice guide with traits, word list and mechanics, plus a rewritten draft with the changes explained.

How it works

1. Read for patterns, not adjectives Measure what the samples actually do: average sentence length, how often they use "you" vs "we", contractions, questions, numbers, jargon, exclamation marks, how sentences open. Write the numbers down; "short sentences" means nothing until you say "12 words on average".
2. Place the samples on the 4 tone dimensions Nielsen Norman Group's dimensions: formal vs casual, serious vs funny, respectful vs irreverent, matter-of-fact vs enthusiastic. Say where the samples sit on each; NN/g warns against the extreme end of every dimension.
3. Name 3 to 4 traits Each trait is a "this, not that" pair, the form Mailchimp's guide uses ("smart but not snobbish"): "Direct, not blunt", "Expert, not academic". Our default is 4 at most, because nobody remembers more. If two traits produce the same edits, merge them.
4. Prove each trait with a do and a don't The "do" is a sentence quoted from the samples. The "don't" is the same idea written the wrong way. If you can't quote a sample for a trait, drop the trait.
5. Build the word list Words they use (product terms, verbs they repeat), words they avoid (jargon, hype, words the samples never use where competitors do). 8 to 15 entries per side. Include how they name the product and the customer.
6. Write the mechanics and the tone shifts Headline case, numerals vs words, Oxford comma, emoji, exclamation marks, contractions, greeting and sign-off. Only list what the samples settle; mark the rest "not decided". Add 1 line on how tone shifts for bad news (an outage, a price rise): same traits, less humor, more plain.
7. Rewrite the draft Keep every fact and claim; change only wording, order, and length. Then list the 3 to 5 biggest changes with the trait each one serves, so the writer learns the guide instead of just getting new copy.
8. Flag what voice can't fix If the draft lacks proof, a clear offer, or a reader, say so in one line. Don't hide a weak message under a good tone.

The SKILL.md

The whole file, as your AI reads it. Download it

---
name: brand-voice
description: Extracts a brand voice guide from 3 to 5 sample texts (traits with do and don't pairs, a word list, sentence habits) and rewrites a draft in that voice. Use when someone asks for a brand voice, tone of voice guide, style guide, "make this sound like us", "rewrite in our voice", or wants copy from several writers to sound like one company.
---

# Brand voice guide and rewrite

A voice guide is useful only if two different writers who follow it produce copy a reader can't tell apart. So every trait comes with a do and a don't taken from real sentences, and the guide is proven by rewriting one draft in it. Voice is the personality that stays the same; tone is how it shifts with the reader's situation (Mailchimp's style guide: "the same voice all the time, but your tone changes").

## When to use

- A team has a few pieces of copy it likes and wants new copy to match them.
- Writers, freelancers, or AI drafts sound like different companies.
- Someone shares a draft and asks to "make it sound like us".
- Not for inventing a brand from nothing with no samples: ask for 3 texts the team likes, even from other companies.

## What you need

- 3 to 5 sample texts the team says sound right: web pages, emails, posts, support replies. Fewer than 3: ask once for more; if none come, proceed and mark the guide "[assumption: based on 2 samples]".
- Optional: 1 text that sounds wrong to them. A negative sample sharpens every "don't".
- Optional: a draft to rewrite. If none, write the guide only.
- Who the reader is (buyer role, how technical). If missing, infer it from the samples and say so.

## Steps

1. **Read for patterns, not adjectives.** Measure what the samples actually do: average sentence length, how often they use "you" vs "we", contractions, questions, numbers, jargon, exclamation marks, how sentences open. Write the numbers down; "short sentences" means nothing until you say "12 words on average".
2. **Place the samples on the 4 tone dimensions.** Nielsen Norman Group's dimensions: formal vs casual, serious vs funny, respectful vs irreverent, matter-of-fact vs enthusiastic. Say where the samples sit on each; NN/g warns against the extreme end of every dimension.
3. **Name 3 to 4 traits.** Each trait is a "this, not that" pair, the form Mailchimp's guide uses ("smart but not snobbish"): "Direct, not blunt", "Expert, not academic". Our default is 4 at most, because nobody remembers more. If two traits produce the same edits, merge them.
4. **Prove each trait with a do and a don't.** The "do" is a sentence quoted from the samples. The "don't" is the same idea written the wrong way. If you can't quote a sample for a trait, drop the trait.
5. **Build the word list.** Words they use (product terms, verbs they repeat), words they avoid (jargon, hype, words the samples never use where competitors do). 8 to 15 entries per side. Include how they name the product and the customer.
6. **Write the mechanics and the tone shifts.** Headline case, numerals vs words, Oxford comma, emoji, exclamation marks, contractions, greeting and sign-off. Only list what the samples settle; mark the rest "not decided". Add 1 line on how tone shifts for bad news (an outage, a price rise): same traits, less humor, more plain.
7. **Rewrite the draft.** Keep every fact and claim; change only wording, order, and length. Then list the 3 to 5 biggest changes with the trait each one serves, so the writer learns the guide instead of just getting new copy.
8. **Flag what voice can't fix.** If the draft lacks proof, a clear offer, or a reader, say so in one line. Don't hide a weak message under a good tone.

## Output format

```markdown
# [Brand] voice guide

**Reader:** [who we write for, in one line]
**Based on:** [list of samples]

## Traits
| We are | Not | Do (from samples) | Don't |
|---|---|---|---|
| [trait] | [overdone version] | "[quoted sentence]" | "[same idea, wrong]" |

## Sentence habits
- Average sentence: [n] words. Longest allowed: [n].
- [you/we ratio, contractions, questions, numbers]

## Words
| Use | Avoid |
|---|---|
| [word] | [word] — [why] |

## Tone dimensions
Formal–casual: [where] · Serious–funny: [where] · Respectful–irreverent: [where] · Matter-of-fact–enthusiastic: [where]

## Mechanics
- [headline case, numerals, punctuation, emoji, sign-off; "not decided" where samples don't settle it]
- Bad news: [how tone shifts]

## Rewrite
**Before:** [draft]
**After:** [rewrite]

**What changed**
1. [change] — [trait]

**Voice can't fix:** [missing proof/offer, or "nothing"]
```

## Example

**Input:** 3 samples from Fernway, a payroll app for restaurants: the home page, an onboarding email, and 1 support reply. Lines from them: "Payday shouldn't eat your Sunday." "Run payroll in 4 minutes from your phone." "Tips split by hours worked, automatically." Counted across the full samples: sentences average 9 words, the longest has 18, and "you" appears 3 times as often as "we". All 3 use contractions, sentence-case headlines, and numerals for numbers, with no emoji and no questions in headlines. Draft to rewrite: "Fernway is a comprehensive, innovative payroll solution that empowers restaurant owners to streamline their operations."

**Output:**

```markdown
# Fernway voice guide

**Reader:** Restaurant owners and managers, not finance people [assumption: inferred from the samples; confirm with the team].
**Based on:** home page, onboarding email, 1 support reply

## Traits
| We are | Not | Do (from samples) | Don't |
|---|---|---|---|
| Plain | Dumbed down | "Run payroll in 4 minutes from your phone." | "Process compensation disbursements efficiently." |
| On their side | Chummy | "Payday shouldn't eat your Sunday." | "Hey friend! We've got your back!!" |
| Specific | Number-stuffed | "Tips split by hours worked, automatically." | "Advanced tip allocation capabilities." |

## Sentence habits
- Across the full samples, not just the lines quoted above: average sentence 9 words; longest allowed 18.
- "You" appears 3 times as often as "we". Contractions always. No questions in headlines.

## Words
| Use | Avoid |
|---|---|
| payday, run payroll | solution, operations — say nothing |
| tips, split | empower, streamline — hype verbs |
| hours worked, automatically | comprehensive, innovative — claims with no proof |
| from your phone, you (never "restaurant owners") | compensation, capabilities — finance and software jargon |

## Tone dimensions
Fairly casual · Mostly serious, light humor in headlines · Respectful · Matter-of-fact

## Mechanics
- Sentence case headlines. Numerals for all numbers. No emoji. Oxford comma: not decided.
- Bad news: drop the humor, keep the short sentences.

## Rewrite
**Before:** Fernway is a comprehensive, innovative payroll solution that empowers restaurant owners to streamline their operations.
**After:** Fernway runs payroll for your restaurant, start to finish.

**What changed**
1. "Payroll solution" became what it does: "runs payroll" — Plain
2. "Restaurant owners" became "your restaurant" — On their side
3. "Comprehensive" became "start to finish"; "innovative" and "streamline their operations" were cut, as they claim nothing a reader can check — Specific

**Voice can't fix:** proof. The draft gives no number. The samples' "Run payroll in 4 minutes from your phone." would fit if it holds for this page [check: the 4-minute claim applies here].
```

## Checks before you send

- [ ] Every "do" is a sentence quoted from a sample, not written by you.
- [ ] 3 or 4 traits, each a "this, not that" pair.
- [ ] Sentence habits include at least 2 measured numbers, and all 4 tone dimensions are placed.
- [ ] The rewrite keeps every fact in the draft and adds none.
- [ ] Each listed change names the trait it serves.

## Common mistakes

- Traits like "friendly, professional, innovative". Every company claims them; they change no sentence.
- Rewriting into your own house style instead of the samples'. Recheck sentence length against the measured average.
- Adding claims during the rewrite to make it punchier. Voice changes wording, never facts.

Share it with your team

Add it to an AI Skill Pod and everyone on the team, contractors included, uses the same version from their own AI. Read the best ways to share AI skills with a team

Add to my team's pod

Changelog

• v1.0.2 · 4 November 2026 · Example: the guide's sentence habits now say the 9-word average and 18-word longest were counted across the full samples, not the 3 lines it quotes.
• v1.0.1 · 4 November 2026 · Example: the input now names the 3 samples, quotes all 3 lines the guide uses, and gives the counted sentence habits and mechanics the guide reports; the reader is marked as an assumption, the word list has 8 entries a side from the input, and the rewrite keeps the draft's claims and adds none, with the samples' 4-minute claim offered as a check.
• v1.0.0 · 4 November 2026 · First release.