QuQi

Build a Voice Guide From Published Work

Voice guides are usually a list of adjectives - clear, confident, human, approachable - which every writer already believes describes them, so no draft changes. The useful artefact is descriptive rather than aspirational: it names the patterns in the pages you already consider good and shows the rejected alternative beside each one. Build it once you have enough published work to sample from, and rebuild it when the people doing the writing turn over.

Get the skill file Let the agents run it
CATEGORY
Content production
FORMAT
voice-guide-from-published-work.md
STEPS
7
PRICE
Free - no account
WHEN TO REACH FOR THIS

Use when several writers are producing pages that each read differently and nobody can state concretely what the house voice is.

The skill file

voice-guide-from-published-work.md
---
name: voice-guide-from-published-work
description: Use when several writers are producing pages that each read differently and nobody can state concretely what the house voice is.
---

# Build a Voice Guide From Published Work

Voice guides are usually a list of adjectives - clear, confident, human, approachable - which every writer already believes describes them, so no draft changes. The useful artefact is descriptive rather than aspirational: it names the patterns in the pages you already consider good and shows the rejected alternative beside each one. Build it once you have enough published work to sample from, and rebuild it when the people doing the writing turn over.

## What you need first

- ten to fifteen published pages the team agrees are good, picked before any analysis
- three or four pages the team considers off-voice, which are what make the rules legible
- the actual editing notes from recent review rounds, which usually contain the rules already
- a decision on who holds final say on voice, or the guide gets argued with instead of used

## Method

1. Collect the good and the bad samples first and do not justify the choices yet. Choosing by feel and then finding the pattern is more honest than deciding a pattern and hunting for evidence of it.
2. Read the openings of all the samples side by side. Openings carry most of the perceived voice, and the difference between the two piles is usually visible within two sentences.
3. Extract mechanical facts rather than impressions: typical sentence length, whether you address the reader as you, whether contractions appear, whether headings are questions or noun phrases. These are checkable by a writer at midnight; "confident" is not.
4. Mine the last few rounds of editing notes for repeated corrections. Any note an editor has written three times is already a rule and belongs in the guide.
5. Write each rule as a pair: one sentence from a real published page, and the version you rejected. Paired examples change drafts, adjectives do not.
6. Cap the guide at two pages and put the three most broken rules at the top. A long guide is read once during onboarding and never opened again.
7. Add a short banned list of constructions you will always cut, so writers stop producing them rather than editors removing them every time.

## What this produces

A two-page descriptive voice guide with mechanical rules, paired before-and-after examples and a banned-construction list.

## Where this goes wrong

- Writing the guide from how you wish you sounded rather than from pages you actually published, producing a document no draft ever matches
- Filling it with format rules such as heading levels and image sizes, which buries the writing guidance readers came for
- Freezing it, so it enforces the habits of writers who left and obstructs the ones who arrived

---

From the QuQi skill library - https://www.quqi.io/skills/voice-guide-from-published-work
Free to download · no account, no email

What you need first

  • ten to fifteen published pages the team agrees are good, picked before any analysis
  • three or four pages the team considers off-voice, which are what make the rules legible
  • the actual editing notes from recent review rounds, which usually contain the rules already
  • a decision on who holds final say on voice, or the guide gets argued with instead of used

Method

  1. 01 Collect the good and the bad samples first and do not justify the choices yet. Choosing by feel and then finding the pattern is more honest than deciding a pattern and hunting for evidence of it.
  2. 02 Read the openings of all the samples side by side. Openings carry most of the perceived voice, and the difference between the two piles is usually visible within two sentences.
  3. 03 Extract mechanical facts rather than impressions: typical sentence length, whether you address the reader as you, whether contractions appear, whether headings are questions or noun phrases. These are checkable by a writer at midnight; "confident" is not.
  4. 04 Mine the last few rounds of editing notes for repeated corrections. Any note an editor has written three times is already a rule and belongs in the guide.
  5. 05 Write each rule as a pair: one sentence from a real published page, and the version you rejected. Paired examples change drafts, adjectives do not.
  6. 06 Cap the guide at two pages and put the three most broken rules at the top. A long guide is read once during onboarding and never opened again.
  7. 07 Add a short banned list of constructions you will always cut, so writers stop producing them rather than editors removing them every time.

What this produces

A two-page descriptive voice guide with mechanical rules, paired before-and-after examples and a banned-construction list.

Where this goes wrong

  • Writing the guide from how you wish you sounded rather than from pages you actually published, producing a document no draft ever matches
  • Filling it with format rules such as heading levels and image sizes, which buries the writing guidance readers came for
  • Freezing it, so it enforces the habits of writers who left and obstructs the ones who arrived

Use this skill in your own AI

The download is a plain markdown file with the name and trigger in its frontmatter. Where an assistant supports skills it can load itself, that frontmatter is what it reads to decide this one applies.

Claude Code Save it as ~/.claude/skills/voice-guide-from-published-work/SKILL.md and Claude loads it on its own when what you are doing matches the trigger line. Put it in .claude/skills inside a project instead if the whole team should have it.
Claude Upload the file in the skills section of your settings. Once it is there it applies itself in any conversation where the trigger fits, so you do not have to remember it exists.
ChatGPT There is no skills format to install into, so paste the file contents into a Project instruction or a Custom GPT instead. It then applies to every chat in that project rather than only the one you paste it into.
Anything else Paste the markdown into the chat before your question. It works in any assistant, it just has to be pasted again each time.

Questions about this skill

When is a voice guide worth building, and why not just list the qualities we want?

When several writers each read differently and nobody can state the house voice concretely. A list of adjectives such as clear, confident and human changes no drafts, because every writer already believes it describes them. Build the descriptive version once you have ten to fifteen published pages worth sampling, and rebuild it when the writing team turns over.

What do I need to gather first, and what happens if I skip part of it?

Ten to fifteen published pages the team agrees are good, three or four it considers off-voice, the actual editing notes from recent review rounds, and a decision on who holds final say. The off-voice pile is what makes the rules legible; without it you describe your own writing in the abstract. Without a named decision-maker the guide gets argued with rather than used.

What does the finished guide contain, and which part changes drafts?

Two pages of mechanical rules with paired examples and a banned-construction list. The pairs are what writers apply: one sentence from a real published page beside the version you rejected. Mechanical facts such as typical sentence length, whether you address the reader as you, and whether headings are questions can be checked by a writer at midnight. Confident cannot.

What ruins a voice guide?

Writing it from how you wish you sounded rather than from pages you actually published, which produces a document no draft ever matches and every editor overrules. Filling it with format rules for heading levels and image sizes does quieter damage: it buries the writing guidance and pushes the guide past the length anyone reopens after onboarding.

More in Content production