Brand Voice

Produce a voice guide: a compact, enforceable document that makes every future piece of copy sound like the same brand, whoever (or whatever) writes it. Voice is how the brand sounds; positioning is what it claims — keep them separate. The governing principle: a voice guide is only real if it changes sentences. Every rule in it must be concrete enough that you can point at a line of copy and say pass or fail.

When to use / when not to

• Use before writing copy at scale, when content across channels sounds inconsistent, or when the user asks for tone/voice/brand guidelines.
• If the user's actual problem is what to say — differentiation, the core claim — use skills/marketing/positioning-angles first; voice without positioning is a costume on an empty claim.
• If the user wants one specific page written, use skills/marketing/direct-response-copy (it consumes the voice guide; a lightweight voice sketch from its own intake is acceptable if no guide exists).
• Audit mode: if a voice guide already exists and the user says "our content drifted", skip to Step 6 and score existing content against the guide.

Intake

Ask in one batch, only for what's missing from context:

1. What do you sell and to whom? (If a positioning statement exists from skills/marketing/positioning-angles, use it — don't re-ask.)
2. Paste 2–3 samples of your current writing (site copy, emails, posts) — or share a URL.
3. Name 2–3 brands or writers whose voice you admire, and one you never want to sound like.
4. Any words your brand insists on or bans?

Don't stall: if samples or a site exist, infer answers 3–4 from them, state your inferences as assumptions, and proceed.

Workflow

1. Extract the current voice (if any content exists)

Before prescribing, describe. Read the samples and write a 3-line diagnosis: what the voice currently is, where it's inconsistent, what's distinctive worth keeping. Skip only for pre-launch brands with no content.

2. Propose an archetype — user decides

Propose the top 2–3 archetype candidates with a one-line case for each; have the user pick one. Never present all ten as a menu.

Decision rules for shortlisting: trust-critical B2B or expertise-led → Sage or Ruler; challenger attacking an incumbent category → Outlaw or Magician; care/health/support products → Caregiver or Innocent; consumer brands built on relatability → Everyman; premium/design-led → Creator or Lover; performance and ambition → Hero. The archetype must match what the brand can sustain, not what sounds coolest.

3. Set the eight voice dimensions

Place the brand on each spectrum (1–5) with a one-line justification tied to the audience or positioning:

Rule: push 2–3 dimensions to an extreme (1 or 5); leave the rest moderate. All-moderate settings produce a generic voice; all-extreme settings produce a caricature. The extremes are where distinctiveness lives — choose the ones the competition can't or won't match.

4. Write the enforcement rules

• Do's / Don'ts — 5 each, all testable against a sentence ("Do open with the reader's problem, not our product"; not "Do be authentic").
• Word choices — words/phrases the brand uses, and banned words (including the category's clichés).
• Sentence structure — length tendency, active/passive, contractions, punctuation habits (em-dashes, exclamation marks), person (we/you/I).

5. Build the tone matrix

Voice is constant; tone flexes by context. For each context, state which dimensions shift and by how much — never the archetype:

Educational content · Sales/promotional · Customer support · Social media · Crisis communication

Example rule format: "Support: Warmth 4→5, Humor 3→1, everything else unchanged."

6. Prove it with rewrites

Take 3 real passages from the user's content (or draft realistic ones if pre-launch) and show before → after in the new voice, each with one line naming which rules drove the change. This is the acceptance test: if the rewrite isn't visibly different and better, the guide's rules are too vague — tighten them and redo.

Required output format

# Voice Guide — [Brand]

**Archetype:** [name] — [one-line why]
**Voice in one sentence:** [e.g., "A sharp, warm expert who talks like a person, not a deck."]

## Dimensions
| Dimension | Setting (1–5) | Why |
|---|---|---|
| Formality | ... | ... |
(all eight; mark the 2–3 extremes in bold)

## Rules
**Do:** 1–5 (testable)
**Don't:** 1–5 (testable)
**Say:** [preferred words/phrases] · **Never say:** [banned words]
**Sentences:** [structure rules]

## Tone matrix
| Context | Dimension shifts | Note |
|---|---|---|
| Educational | ... | ... |
| Sales | ... | ... |
| Support | ... | ... |
| Social | ... | ... |
| Crisis | ... | ... |

## Before / After
1. Before: "..." → After: "..." — [rule applied]
2. ...
3. ...

In audit mode, append: a scorecard of sampled content (pass/fail per rule) and the top 3 drift patterns to fix.

Quality bar (check before delivering)

• [ ] 2–3 dimensions at an extreme, justified; not all-moderate, not all-extreme.
• [ ] Every do/don't is checkable against a single sentence of copy — no "be authentic"-grade rules.
• [ ] At least 3 before/after rewrites, using the user's real content when any exists.
• [ ] Swap test: replace the brand name with its nearest competitor's — if the guide still fits them, it's too generic; sharpen the extremes and word choices.
• [ ] Tone matrix changes dimensions only, never the archetype.
• [ ] No claims about the brand's audience invented — inferences labeled as assumptions.

Integration

• Upstream: skills/marketing/positioning-angles supplies the positioning statement — the claim the voice must carry. skills/marketing/orchestrator routes here as a foundation step.
• Downstream: the voice guide is a required input to skills/marketing/direct-response-copy, skills/marketing/seo-content, skills/marketing/email-sequences, skills/marketing/newsletter, skills/marketing/content-atomizer, and skills/marketing/tweet-writer. What crosses the boundary is this document — paste or reference it whenever any of those skills write copy.