Brand Voice
Design the voice an organization should use. Turn brand strategy into concrete writing rules that a person or agent can apply and review.
Boundary
Use this skill to define a desired organizational voice.
- To capture how a person or organization already writes from real samples, use
voice-dna-kit. - To measure a sample's style deeply, use
style-forensics. - To draft content from an existing voice profile, use
style-writer.
Strategic market identity, differentiation, and the claims the voice expresses are decided upstream and are not set here.
Existing brand content may reveal current habits, but do not treat those habits as the desired voice automatically.
Inputs
Collect the smallest useful set:
- Brand positioning and product promise.
- Primary audiences and what they need to understand, feel, or do.
- Current perception and desired perception.
- Three to five traits the voice should express.
- Important channels and contexts.
- Examples the organization admires or rejects, when available.
- Drafts paired with versions approved or edited by the brand owner, when available.
- Legal, accessibility, terminology, or claims constraints.
If positioning or audience is unresolved, flag that gap. Voice rules cannot repair unclear strategy.
Workflow
1. Assess source quality
Label each input by owner, channel, audience, date, approval status, and relevance to the desired positioning. Separate:
- approved examples that represent the desired voice
- current material that reveals habits but not necessarily intent
- drafts and their owner-approved edits
- admired external examples used only to identify qualities or boundaries
- outliers, mixed-author work, and material from obsolete positioning
Do not average conflicting sources into a fictional voice. State which evidence governs and which material was excluded or treated as provisional.
2. Define the voice's job
State:
- who the brand is speaking to
- what the audience should perceive
- what must remain true in every context
- what the brand must never sound like
Separate identity from tone. Identity stays stable. Tone changes with context.
3. Turn traits into observable behavior
Choose three to five traits. Define each trait with:
- what it means
- how it appears in writing
- what it does not mean
- one positive example
- one negative example
Reject unsupported adjective lists such as "bold, authentic, innovative." A useful trait must change a writing decision.
4. Define constants and tone dials
Document the characteristics that never change, then specify how tone shifts across contexts such as:
- marketing
- sales
- onboarding
- support
- incidents and errors
- product interfaces
- legal or policy communication
For each context, name the purpose, tone adjustment, evidence requirement, and prohibited behavior.
5. Set expression rules
Define concrete guidance for:
- vocabulary and terminology
- sentence and paragraph rhythm
- formality and contractions
- confidence and hedging
- humor and emotional warmth
- pronouns and audience address
- punctuation and formatting
- claims, proof, and specificity
- openings, transitions, and closings
Prefer enforceable rules over mood words.
6. Build calibration examples
Create paired off-voice and on-voice examples for at least three important contexts. Change only what the guide explains so reviewers can trace each revision to a rule.
Do not copy admired brands. Use them only to identify qualities or boundaries.
7. Test the guide
Draft three short messages with different stakes. Review whether:
- the same identity survives across contexts
- tone changes appropriately
- the output is distinguishable from generic category copy
- every correction maps to a written rule
- legal and factual constraints remain intact
Patch the guide where reviewers rely on taste that the guide does not encode.
8. Run owner-edit calibration
When a brand owner supplies an approved edit, compare it with the proposed draft and record:
- exact wording retained, removed, or replaced
- the audience or risk the edit addresses
- the candidate rule the edit suggests
- whether other approved examples support the same rule
One edit creates a candidate preference, not a universal rule. Promote it into the guide only when the owner confirms it or the pattern recurs across approved examples. Record contradictions instead of forcing consensus.
Deliverable
Use brand-voice-guide.md as the structure. Save the result to the location requested by the user. If no location was requested, present the guide before creating a file.
The guide should contain:
- Brand and audience context
- Core voice principles
- Trait definitions with boundaries
- Voice constants and contextual tone matrix
- Language, rhythm, punctuation, and formatting rules
- Claims and evidence rules
- Off-voice and on-voice examples
- Owner-edit calibration log and confirmed rules
- Review checklist
- Open assumptions and revisit triggers
Quality Gate
The guide is ready when:
- each trait changes observable writing behavior
- context changes tone without changing identity
- examples demonstrate rather than restate rules
- banned patterns are concrete
- reviewers can explain why a draft passes or fails
- Owner-edit calibration distinguishes one-off preferences from repeated or confirmed rules
- strategic assumptions and unresolved gaps are explicit