Agent skill · Writing

Brand voice profile

Builds a reusable voice guide from 3 to 10 writing samples: sentence length, words to use and avoid, formatting habits, openings and sign-offs, and do and don't pairs quoted from the samples. Saves it as VOICE.md so other skills and prompts can follow it. Use when asked to write in someone's voice, match a brand's tone, or set up a style guide from examples.

The brand-voice-profile/ folder with SKILL.md and its reference files.

The skill itself. Agents read the name and description first and load the rest only when a task matches.

brand-voice-profile/SKILL.md
---
name: brand-voice-profile
description: "Builds a reusable voice guide from 3 to 10 writing samples: sentence length, words to use and avoid, formatting habits, openings and sign-offs, and do and don't pairs quoted from the samples. Saves it as VOICE.md so other skills and prompts can follow it. Use when asked to write in someone's voice, match a brand's tone, or set up a style guide from examples."
---

# Brand voice profile

Describe how someone writes, from their own words, so another writer (or agent) can match it.

## Steps

1. Collect 3 to 10 samples the person wrote themselves: posts, emails, docs. Skip anything edited by someone else or written by AI. With fewer than 3, or under 300 words in total, say the profile is low-confidence.
2. Measure: average sentence length and how much it varies, paragraph length, use of lists, headings, emoji, exclamation marks, capitals and line breaks.
3. Read for habits: how pieces open (quote three real first lines), how they close, words and phrases they repeat, words they never use, how they say no, how certain they sound, how they refer to themselves and the reader.
4. Write VOICE.md using [references/VOICE-TEMPLATE.md](references/VOICE-TEMPLATE.md): 8 to 12 rules, each with a quote from the samples as evidence, then do and don't pairs, then a banned list.
5. Test it: rewrite one neutral paragraph in the voice, and show it next to the original.

## Rules

- Every rule needs a quote from the samples. No quote, no rule.
- Describe, don't flatter: "short declarative sentences, rarely over 15 words", not "punchy and engaging".
- Start the banned list from common AI phrasing (delve, game-changer, unlock, in today's fast-paced world, it's not just X it's Y) and add anything that would sound wrong in this voice.
- Keep it under 400 words so it fits in another prompt.