---
name: brand-voice-profile
source: https://app.decimal.ai/s/brand-voice-profile@1/SKILL.md
source_sha256: 27dea72b6418
---

# Brand Voice Profile — output convention

## Contract

When asked to analyze, extract, or document a company's brand voice from its
content, emit a Markdown document in EXACTLY the fixed Brand Voice Profile
template below — fixed section names, fixed order, fixed table shape, fixed
labels. Apply whenever the deliverable is a voice/tone profile, guide, or
reference a writer will follow. Never answer with free-form prose analysis.

## Rules (the complete template)

**R1 — Title.** First line is exactly `# Brand Voice Profile: <Company Name>`.
Use the company name verbatim, including punctuation (`Atlas Moving Co.`,
`Kettle & Co`, `Sprout & Spore`).

**R2 — Metadata block.** Immediately under the title, two lines:
- `**Analyzed:** <date> | **Content pieces analyzed:** <N>` (a real count, not a range)
- `**Sources:** <list of the analyzed pieces>`

**R3 — Exactly six `##` sections, in THIS order, with THESE exact names:**
1. `## Voice Summary`
2. `## Tone Profile`
3. `## Language & Vocabulary`
4. `## Structure & Formatting`
5. `## Audience & Persona`
6. `## Writing Guidelines (Actionable)`

Do not add, drop, rename, reorder, or merge sections. The parenthetical
`(Actionable)` is part of the sixth heading.

**R4 — Voice Summary.** 2 to 3 sentences only. No bullets, no subsections.

**R5 — Tone Profile is a Markdown table.** Columns are exactly, in order:
`Dimension | Position | Evidence`. The table has exactly these five rows, in
this order: **Formality**, **Emotional Register**, **Authority**, **Humor**,
**Directness**. No extra rows, no missing rows, no renamed dimensions.

**R6 — Positions are word spectrums, never numbers.** Each `Position` cell is a
descriptor on a verbal spectrum (e.g. `Casual ↔ Professional ↔ Academic`,
`Measured, occasionally excited`). Never a numeric score, a 1–10 rating, or stars.

**R7 — Language & Vocabulary subsections.** Five `###` subsections, in this order:
`Reading Level`, `Signature Phrases`, `Jargon & Technical Depth`,
`Words They Love`, `Words They Avoid`.

**R8 — Use the labels verbatim.** `Words They Love` and `Words They Avoid` are
literal. Do not substitute `Preferred words`, `Words to use`, `Avoided words`,
or any synonym.

**R9 — Audience & Persona includes Point of View.** This section must contain a
`Point of View` subsection identifying I / we / you / third-person usage.

**R10 — Writing Guidelines block.** `## Writing Guidelines (Actionable)` contains,
in order, a `### Do` list, a `### Don't` list, and a `### Voice Samples` block.
The Voice Samples block has two verbatim-labeled parts: `**Their style:**`
(real quotes from their content) and `**How to match it:**` (new sentences
written in their voice on a neutral topic).

## Worked examples (BEFORE = base default, AFTER = conforming)

**R1 title.**
BEFORE: `# Brand Voice Analysis for Acme` or `## Acme Voice Guide`
AFTER: `# Brand Voice Profile: Acme Robotics`

**R2 metadata.**
BEFORE: (jumps straight from title into prose)
AFTER:
```
# Brand Voice Profile: Acme Robotics
**Analyzed:** 2026-06-26 | **Content pieces analyzed:** 14
**Sources:** acme.com/blog/scaling, acme.com/customers/northwind, ...
```

**R3 / R4 section set + Voice Summary.**
BEFORE: `## Overview\nAcme's tone is friendly and confident across a few paragraphs...`
AFTER: `## Voice Summary\nAcme writes in a confident, peer-to-peer voice for working engineers. It assumes deep technical fluency and favors short, declarative claims over hedging.`

**R5 Tone Profile shape.**
BEFORE:
```
Tone: friendly, professional, a bit playful, authoritative.
```
AFTER:
```
## Tone Profile
| Dimension | Position | Evidence |
|-----------|----------|----------|
| Formality | Professional-casual | "Let's just ship it." |
| Emotional Register | Measured, occasionally excited | "We were genuinely floored." |
| Authority | Expert/teacher | "Here's what most teams get wrong." |
| Humor | Rare, dry | "Yes, it's another dashboard." |
| Directness | Very direct, bold claims | "This is the fastest path. Full stop." |
```

**R6 positions.**
BEFORE: `| Formality | 7/10 | ... |`
AFTER: `| Formality | Casual ↔ Professional (leans professional) | ... |`

**R7 / R8 vocabulary labels.**
BEFORE: `### Preferred Words` ... `### Words to Avoid`
AFTER: `### Words They Love` ... `### Words They Avoid`

**R9 Point of View.**
BEFORE: (Audience section only describes the reader, omits POV)
AFTER:
```
## Audience & Persona
### Point of View
First-person plural ("we") for the brand, second person ("you") for the reader — a peer-to-peer relationship, not institutional.
```

**R10 Voice Samples.**
BEFORE: `Example sentences: ...` (one undivided list)
AFTER:
```
### Voice Samples
**Their style:**
> "You don't need a bigger team. You need a tighter loop."
**How to match it:**
> Stop adding headcount to a process that's leaking time. Fix the loop first.
```

## Edge cases & exceptions

- **Few sources / thin content.** Still emit all six sections. Fill the count
  honestly (`Content pieces analyzed: 4`) and note low confidence inside
  `Voice Summary` — do not drop sections for lack of material.
- **A dimension has no evidence.** Keep the row; put the inferred position and
  write `(no clear signal)` in the Evidence cell. Never delete one of the five rows.
- **Multiple voice modes** (e.g. playful blog, formal docs). Keep one Tone table
  with the dominant position per row; note the split in `Structure & Formatting`
  or `Voice Summary`. Do not add a second Tone table or extra rows.
- **Company name has punctuation/ampersand** (`Kettle & Co`, `Atlas Moving Co.`).
  Reproduce it exactly in the title; do not normalize or strip it.
- **User asks for "a guide" / "a reference" / "not an essay."** That is still this
  template — the profile IS the guide. Do not switch to prose.
- **No real quotes available.** `**Their style:**` should paraphrase closely and
  flag it; never fabricate a verbatim quote. `**How to match it:**` is always new
  original sentences in their voice.

## Do / Don't

- DO emit all six sections in the fixed order. DON'T add an `## Introduction`,
  `## Methodology`, or `## Conclusion`.
- DO render Tone Profile as a five-row table. DON'T render tone as a prose
  paragraph or a bullet list.
- DO use word-spectrum positions. DON'T use 1–10, percentages, or stars.
- DO write the labels `Words They Love` / `Words They Avoid` verbatim. DON'T
  rephrase them.
- DO split Voice Samples into `**Their style:**` and `**How to match it:**`.
  DON'T merge them into one block.
- DO keep Voice Summary to 2–3 sentences. DON'T let it sprawl into paragraphs.
- DO include a `Point of View` subsection. DON'T fold POV into a sentence elsewhere
  and skip the subsection.

## Common mistakes (base defaults to avoid)

1. Free-form prose analysis instead of the template — the single biggest miss.
2. Renaming sections (`## Overview`, `## Tone`, `## Vocabulary`,
   `## Recommendations`) instead of the exact six names.
3. Numeric tone scores (`8/10`) instead of word spectrums.
4. Tone as a bullet list instead of a `Dimension | Position | Evidence` table.
5. Dropping or reordering tone rows (missing Humor, or Directness before Authority).
6. Synonym labels (`Preferred Words`, `Avoid List`) instead of the literal ones.
7. Omitting the `Point of View` subsection.
8. One undivided "examples" block instead of `**Their style:**` + `**How to match it:**`.
9. Letting Voice Summary run long instead of 2–3 sentences.
10. Skipping the metadata line (date + `Content pieces analyzed:` + Sources).

## Quick checklist

- [ ] `# Brand Voice Profile: <Company>` title + metadata + Sources line
- [ ] Six sections, exact names, exact order
- [ ] Voice Summary = 2–3 sentences
- [ ] Tone table: `Dimension | Position | Evidence`, five rows in order
- [ ] Positions are word spectrums, not numbers
- [ ] Vocabulary subsections incl. literal `Words They Love` / `Words They Avoid`
- [ ] Audience & Persona has a `Point of View` subsection
- [ ] Writing Guidelines: `### Do`, `### Don't`, `### Voice Samples` split into
      `**Their style:**` and `**How to match it:**`
