Skill · Brand voice

Brand voice.

Voice is constant. Tone adapts. The doc that distinguishes the two is the one writers actually use.

This skill produces a standalone voice document that anyone (writer, designer, founder, AI assistant) can apply consistently. The four layers stack: attributes, tone shifts, vocabulary and grammar, paired examples. Each layer constrains the one below it. The example library is what gets opened most days.

Set tone firstRead SKILL.md

What this skill produces

A document writers actually open.

The output is a markdown document, typically saved as voice.md, with the four layers and an anti-pattern list of specific phrases or constructions the brand will not use. The doc is reference material for the project's life, not a one-time deliverable. It gets opened in the CMS, in the brief template, in the AI assistant prompt. A doc that does not get used is a doc that does not exist.

The showcase demonstrates voice across the four tone positions. Hewn (skincare) sits at Conversational: warm, ingredient-literate, second-person comfortable, no exclamation marks. Pulse (music) sits at Playful: imaginative product naming, inviting copy, the brand can joke because the rest of the work earns it. Vector (offensive security) sits at Provocative: pointed, operator-grade, willing to alienate buyers who are not the audience. Estate (luxury real estate) sits at Professional: measured, low-register, the property does the talking.

The framework

Four layers, stacked.

Each layer constrains the one below it. Pick the attributes first; they govern the tone shifts; the tone shifts govern vocabulary and grammar; vocabulary and grammar govern the examples.

1. Voice attributes

The constants. The personality traits that define how the brand sounds across every context. Pick three to five. Pair each with what it is not, the failure mode if overdone. Confident, not arrogant. Direct, not blunt. Warm, not saccharine. Witty, not sarcastic. The pairings are not a menu; generate the brand's own. The rejection is what saves writers from overshooting.

Test: If another brand in the same category could honestly claim the same three attributes, the attributes are too generic. Replace with attributes that genuinely distinguish.

2. Tone shifts by context

Voice is constant; tone adapts. Map the major contexts the brand writes in. Onboarding gets warmer and slightly slower-paced. Hero copy runs the signature voice fully on. Error messages stay calm and matter-of- fact, no apology theater. Account deletion gets quiet and respectful, no jokes. Crisis communication is factual and accountable. Pricing is direct and confidence-inspiring. Each context gets one or two sentences in the doc explaining the tone shift.

Test: Read the brand's last error message aloud. Does it sound like the same brand the homepage hero sounds like, or does it sound like a default framework string? If the latter, the tone shift for error messages was never designed.

3. Vocabulary and grammar

The granular dial settings. Vocabulary preferences document what the brand uses, what it avoids, what it has redefined (a SaaS product calling its features “huddles” instead of “meetings”, for example), and whether industry jargon stays (signals expertise) or gets stripped (signals approachability). Grammar rules cover contractions, default sentence length, punctuation marks favored or avoided, pronouns, capitalization style, number formatting, the Oxford comma, and active versus passive voice. Document the choices that genuinely distinguish the brand. Skip the default rules unless the brand has an actual position on them.

Em-dash discipline: The em dash is famously polarizing. Pick a position (use freely, use sparingly, do not use). On RampStack pages we do not use em dashes; the rule lives in this doc and propagates through every downstream skill.

4. Paired-example library

Voice is taught through examples, not rules. Build a library of fifteen to twenty-five paired examples: a bad version (off-voice), a good version (on-voice), and a brief note on what changed. Cover the content types the brand actually writes: headlines, subheadlines, hero CTAs, feature descriptions, testimonial intros, email subject lines, error messages, success messages, about-page paragraphs, social posts. The examples that show nuanced shifts beat the examples that show cartoonish before-and-after.

Test: Pick a fresh writing brief and try to apply the doc cold. If the doc produces on-voice copy, the doc is complete. If it does not, the doc is missing examples for the kind of writing this brief requires.

Composition

How the voice doc fits the rest.

Brand voice consumes the tone axis from creative direction. If the brief says Conversational, the voice doc operationalizes Conversational into named attributes, vocabulary, grammar, and examples. If the brief says Professional, the doc produces a different attribute set. Without the brief, the voice doc gets written against an imagined brand and tends to land on the comfortable middle (friendly, professional, approachable, helpful) that distinguishes nothing.

The voice doc is then required reading for every downstream copy skill. Landing-page copy applies the voice to a specific conversion context. Content-and-copy applies it to long-form editorial. Email-sequences applies it to lifecycle and broadcast sends. The voice stays constant; tone shifts handle the context. A doc that does not get referenced produces inconsistent voice across surfaces, which reads as a brand that has not yet decided who it is.

Voice also feeds into brand style guide. The guide is the central artifact that documents the full brand system; voice is one section of it. The voice doc can stand alone or feed into the guide. Most teams start with the standalone voice doc and incorporate it into the guide once the rest of the system stabilizes.

Failure patterns

Where voice docs go wrong.

Generic attributes. Friendly, professional, approachable, helpful. Every brand says this. Means nothing. Pick attributes that genuinely distinguish, even if they are uncomfortable. A brand that says “Direct, even when it costs us a deal” gives writers something to work with that “professional” never will.

No “we are NOT” pairings. Without the rejection, attributes drift toward extremes. The pairing is the part of the doc that actually does work in the writer's head when they are deciding between two phrasings.

Voice doc with no examples. Rules without examples cannot be applied. Writers cannot infer voice from abstract attributes; they need to see what on-voice copy looks like in the same content type they are about to write.

Examples that are obviously bad and obviously good. Real voice work shows nuanced shifts, not cartoonish before and after. A bad example that no writer would actually produce trains nothing. The useful examples are the ones where both versions sound plausible, but only one is on-voice.

Skipping tone shifts. Treating voice as one-size-fits-all leads to a brand that sounds wrong in error states or legal contexts. The tone-shift section costs an hour to write and saves quarters of inconsistency.

Voice without distribution. A perfect doc that no one references is worth nothing. Make the doc easy to reference inline: link from CMS templates, link from the brief template, paste relevant excerpts into AI assistant prompts. The doc only works if it travels.

Reference

Source files and related skills.

Frequently asked questions.

What does a complete brand voice document include?
Four layers. First, voice attributes paired with what they are not (the failure mode if overdone). Second, tone shifts by context (a list of the major contexts the brand writes in and how voice expresses differently in each). Third, vocabulary preferences and grammar rules (what the brand says, what it avoids, how it punctuates and capitalizes). Fourth, a paired-example library of fifteen to twenty-five real on-voice and off-voice examples with notes on what changed. The example library is the most-used part in practice.
How does brand voice relate to creative direction?
The tone axis from creative direction sets the register; brand voice operationalizes that register into specific attributes, vocabulary, grammar, and examples. A brief that picks 'Conversational' produces a voice doc that names contractions as default, short paragraphs, second-person comfortable, and a list of the specific marketing cliches the brand will not use. A brief that picks 'Professional' produces a different doc. Run creative direction first; consume its output here.
Why are 'we are X, not Y' pairings important?
Without the rejection, attributes drift toward extremes. 'Confident' alone produces swagger; 'confident, not arrogant' tells writers where the line is. 'Warm' alone produces saccharine; 'warm, not saccharine' anchors the warmth. The pair is what saves writers from overshooting. Generic attribute lists ('friendly, professional, approachable') belong to every brand and distinguish none. The pairings are where the voice doc earns its keep.
What is the most common voice failure pattern?
Documenting an aspirational voice the brand does not actually sound like and has no plan to shift toward. The doc reads beautifully and produces nothing because writers cannot apply guidance the existing copy contradicts. A useful voice doc reflects the actual current voice with a clear delta to where the brand is going, including the editorial cleanup work that delta requires.