Brand Voice Extractor
Extract or build a brand voice profile for consistent, on-brand AI content generation.
Installation
- Make sure Claude is on your device and in your terminal.
Skills load from
~/.claude/skills/when Claude Code starts up β so you need it on your machine first. If you don't have it yet, install it once with the command below, then runclaudein any terminal to verify.One-time setupnpm i -g @anthropic-ai/claude-codeAlready have it? Skip ahead.
- Paste into Claude Code or into your terminal.
This copies the whole skill folder into
~/.claude/skills/brand-voice-extractor-brianrwagner/β the SKILL.md plus any scripts, reference docs, or templates the skill ships with. Works for every skill.What this command does
- Makes a new, empty temporary folder.
- Downloads the skill's folder,
brand-voice-extractor, fromgithub.com/BrianRWagner/ai-marketing-claude-code-skills(branchmain), the GitHub repository it's listed from. You get the version that is there now, not a copy we keep. - Copies that folder to
~/.claude/skills/brand-voice-extractor-brianrwagner/. - Deletes the temporary folder.
Installing runs none of the skill's files; it only downloads and copies them. When you later use the skill, Claude follows its SKILL.md and may run scripts that come with it. Read them first.
Faster alternative (instruction-only skills)
Skips the clone and grabs only the SKILL.md file. Don't use this if the skill ships Python scripts, reference markdowns, or asset templates β they won't be downloaded and the skill will fail when it tries to load them.
Quick install (SKILL.md only)Sign up to copy - Restart Claude Code.
Quit and reopen Claude Code (or any other agent that loads from
~/.claude/skills/). New skills are picked up on startup. - Just ask Claude.
Skills auto-activate when your request matches the skill's description β no slash command needed. Trigger phrases live in the skill's own frontmatter; you can read them in the βWhat this skill doesβ section above.
Prefer to read the source first? Open on GitHub.
When Claude uses it
Extract or build a distinct brand voice profile that AI agents can use to produce on-brand content every time. Two modes: Extract (analyze content you're proud of) or Build (construct a voice from scratch). Outputs a complete voice profile with personality traits, tone spectrum, vocabulary guide, rhythm patterns, and example phrases. Use this before any content creation skill for consistent, human-sounding output.
What this skill does
Brand Voice Extractor
Mode
Detect from context or ask: "Quick snapshot, full Voice Guide, or full guide with AI training examples?"
| Mode | What you get | Best for |
|---|---|---|
quick | Top 5 voice traits + 3 core do/don't rules | Fast reference before a single piece |
standard | Complete Voice Guide: personality, tone, vocabulary, rhythm, phrases | AI training, ghostwriting, brand docs |
deep | Full Voice Guide + 10 before/after rewrites + AI prompt template + onboarding guide | Content team onboarding, agency handoff |
Default: standard β use quick for a fast reference. Use deep if you're onboarding writers or building a content system.
Runtime Context
Platform: Claude Code
File system: Available. Read prior outputs before starting. Save all outputs to the paths specified in Memory Protocol.
Cross-skill dependencies:
- Upstream: None required β this is the starting point
- Downstream: This skill produces a complete VOICE-PROFILE.md ready for use in any content generation workflow.
Memory Protocol
Save output to: voice/[name]-voice-profile-YYYY-MM-DD.md
At session start: Check if a prior voice profile file exists for this person/brand. If yes:
- Load it into context
- Note this is a re-run / refinement session
- Compare new samples to prior profile β flag if voice has evolved or if patterns conflict
Cross-session rule: If a prior voice profile exists, always load it before starting β do not treat every run as a cold start. Reference what was previously extracted and ask if anything has changed since.
Generic copy converts worse than copy with a distinct voice.
Not because the words are different β because the reader feels like they're hearing from a person, not a marketing team.
This skill defines that voice. Either by extracting it from content you're already proud of, or building it strategically from the ground up. The output is a reusable voice profile that every content skill can reference β so your AI never sounds like everyone else's AI.
What You Get
A complete voice profile document containing:
- Voice summary (2-3 sentences capturing the essence)
- Core personality traits with real-world implications
- Tone spectrum across 5 key dimensions
- Vocabulary guide: words to use, words to kill
- Rhythm & structure patterns
- Example phrases (on-brand vs. off-brand)
- Do's and Don'ts for any writer (human or AI)
Result: Anyone β or any AI agent β can read this profile and produce content that sounds unmistakably like you.
Two Modes
Mode 1: Extract
Use when: You have existing content that already sounds right.
Feed the skill 3-5 pieces of content you're proud of β website copy, emails, posts, newsletter editions, anything where you thought "yes, this is me."
The skill analyzes patterns across tone, vocabulary, rhythm, structure, and POV, then codifies what makes your writing distinctive.
Best inputs:
- Your About page or homepage copy
- Your top-performing social posts
- Emails where you nailed the tone
- Newsletter editions you felt good about
- Video or podcast transcripts
Mode 2: Build
Use when: You're starting fresh, your existing content is generic, or you want to evolve your voice strategically.
The skill asks 10-15 targeted questions about your personality, audience, positioning, and aspirations β then constructs a voice aligned with who you are and who you're talking to.
How to choose:
"Do you have existing content that represents how you want to sound?"
- Yes β Extract mode
- No / Not sure β Build mode
How to Use This Skill
Step 1: Choose your mode
Tell the agent: "Run the brand-voice-extractor in Extract mode" or "Build mode."
Step 2: Provide inputs
- Extract: Paste 3-5 pieces of content
- Build: Answer the 10-15 questions the agent asks
Step 3: Receive your voice profile
The agent outputs a complete, formatted voice profile document.
Step 4: Save it
Store the profile somewhere accessible (e.g., brand/VOICE-PROFILE.md in your workspace). Reference it in every future content task.
Step 5: Reference it in other skills
"Write 5 LinkedIn posts using this voice profile: [paste profile]" "Draft an email sequence. Use my voice profile for tone and style." "Rewrite this landing page to match my voice."
Extract Mode: What the Agent Analyzes
Extraction Reasoning Chain
Before outputting patterns, follow this process explicitly:
Analyze samples β Extract signals per dimension β Synthesize conflicts (when patterns contradict, weight 3 most recent samples 2x) β Validate against quality bar β Output
Do not skip steps. Each stage informs the next. Surfacing the reasoning process catches extraction errors before they lock into the profile.
Input Constraints
- Minimum sample length: If total sample content < 500 words, request more before proceeding. A profile built on too little material will be too general to be useful.
- Corporate/committee voice detection: If samples read as committee-written (passive voice >30%, no personal opinions, heavy hedging throughout), flag this to the user before proceeding: "These samples may reflect corporate/edited voice rather than your authentic voice. For best results, share something you wrote quickly and unfiltered β an email, a Slack message, a quick post."
Tone patterns
- Formal β Casual (contractions? fragments? slang?)
- Serious β Playful (humor? gravity?)
- Reserved β Bold (strong claims vs. hedging?)
- Distant β Intimate (I/you vs. we/they?)
Vocabulary patterns
- Jargon level (heavy, translated, or light)
- Signature words and phrases
- Words you seem to avoid
- Everyday vs. formal vocabulary
Rhythm patterns
Technical HOW: Count average words per sentence across all samples. Measure sentence length variance (do most sentences cluster around the same length, or does it range widely?). Flag if >40% of sentences are similar length β that signals a monotone rhythm that may need to be intentional, not accidental.
- Average sentence and paragraph length
- Mix of short/punchy vs. longer/flowing
- Fragment use
- List frequency
Transition patterns
Technical HOW: Scan for bridge phrases appearing 3+ times across samples. Extract the exact wording β not paraphrases, the literal phrases. These become the transition fingerprints. Examples: "Here's the thing...", "What that means is...", "The reality is..."
- Recurring bridge phrases (extract verbatim)
- How they move between ideas (abrupt? flowing? with a pivot?)
- Signature openers and closers
Structural patterns
- How you open (story? question? bold claim?)
- How you transition between ideas
- How you close (CTA? summary? open loop?)
- Headers, formatting, whitespace use
Personality signals
- Self-deprecating or confident?
- Teacher or peer?
- Polished or raw?
- Optimistic or realistic?
Conflict Resolution
When patterns contradict across samples (e.g., some samples are formal, some are casual):
- Weight the 3 most recent samples 2x in your analysis
- Flag the conflict explicitly in the profile: "Voice is inconsistent across [dimension] β recent samples lean [X], older samples lean [Y]. Profile reflects recent direction."
- Ask the user: "I'm seeing tension in [dimension]. Which direction is closer to where you want to go?"
Build Mode: Questions the Agent Asks
Identity (who you are)
- What are 3-5 words that describe your personality?
- What do you stand for? What's your core belief about your industry?
- What's your background? What shaped how you see things?
- What makes you genuinely different from others in your space?
Audience (who you're talking to) 5. Who are you talking to? (Be specific β not "entrepreneurs") 6. What tone resonates with them? What do they respond to? 7. What would make them trust you? What would turn them off?
Positioning (how you show up) 8. Are you the expert, the peer, the rebel, the guide, the insider? 9. Where do you sit on accessible β exclusive? 10. Where do you sit on approachable β authoritative?
Aspiration (what you want to sound like) 11. Name 2-3 people or brands whose voice you admire. What specifically? 12. What do you NOT want to sound like? 13. Any signature words or phrases that feel like "you"? 14. Any words you hate or want to avoid? 15. How do you feel about humor? Profanity? Hot takes?
Voice Profile Output Format
# [Your Name] Voice Profile
## Voice Summary
[2-3 sentences. What does this voice FEEL like to encounter?]
## Core Personality Traits
- **[Trait]:** [What this means in practice]
- **[Trait]:** [What this means in practice]
- **[Trait]:** [What this means in practice]
## Tone Spectrum
| Dimension | Position | Notes |
|-----------|----------|-------|
| Formal β Casual | [position] | [specifics] |
| Serious β Playful | [position] | [specifics] |
| Reserved β Bold | [position] | [specifics] |
| Simple β Sophisticated | [position] | [specifics] |
| Warm β Direct | [position] | [specifics] |
## Vocabulary
**USE:** [words, phrases, signature openers]
**AVOID:** [words, corporate-speak, AI-sounding phrases]
**Jargon level:** [Heavy / Light / Translated]
## Rhythm & Structure
**Sentences:** [pattern]
**Paragraphs:** [pattern]
**Openings:** [signature moves]
**Formatting:** [headers, bullets, whitespace]
## Example Phrases
**On-brand:** [3 examples]
**Off-brand:** [3 examples + why wrong]
## Do's and Don'ts
**DO:** [3-5 rules]
**DON'T:** [3-5 rules]
Quality Bar
A good voice profile passes this test:
β Recognizable β Could someone identify content as "yours" without a byline? β Actionable β Could a writer (human or AI) produce on-brand content using only this profile? β Differentiated β Does it sound different from competitors? β Authentic β Does it feel true to who you are? β Consistent β Can it apply across formats (social, email, long-form)?
If any answer is no, the profile needs more specificity.
After Delivering the Profile
Once the voice profile is output, always offer the next step:
Your voice profile is complete. What's next?
A) Test it β Generate 3 sample sentences in this voice to validate it feels right
B) Refine it β Tell me what's off; I'll diagnose which pattern needs adjustment
C) Strengthen it β I'll identify the weakest dimension and deepen it with more sample analysis
D) Done β Save to voice profile file; it's ready for use in any content generation workflow
Just reply with a letter or describe where you want to go.
How This Connects to Other Skills
This skill produces a complete VOICE-PROFILE.md ready for use in any content generation workflow.
Voice profile β direct-response-copy: "Write landing page copy using this voice profile." Voice profile β content-atomizer: "Repurpose this using my voice profile for tone." Voice profile β email-sequences: "Draft this sequence. Match my voice." Voice profile β lead-magnet: "Frame the lead magnet to match this voice."
The workflow: Run brand-voice-extractor first β Save the profile β Reference it in everything else.
Related skills
A/B Test Designer
coreyhaines31
Design and plan A/B tests to measure which version performs better.
Ad Creative Generator
coreyhaines31
Generate and iterate high-performing ad copy, headlines, and variations for any platform.
Cold Email Writer
coreyhaines31
Write B2B cold emails and follow-up sequences designed to get replies.
Marketing Copy Editor
coreyhaines31
Review and polish marketing copy to strengthen messaging and improve readability.