Extract someone's authentic writing voice from samples, build a complete Voice DNA profile, then generate content that sounds like them — not AI. Covers confidence calibration, energy mapping, transition patterns, audience adaptation, and platform-specific voice tuning. Triggers on: capture my voice, write like me, voice guide, brand voice, sound like me, voice profile, my writing style, content in my voice, doesn't sound like me, too AI.
openclaw4,189 스타2026. 2. 20.
직업
카테고리
콘텐츠 제작
스킬 내용
The #1 complaint about AI content: "It doesn't sound like me."
This skill fixes that permanently. Not with a one-line tone instruction. With a complete voice operating system that understands HOW someone communicates — their patterns, energy, confidence zones, transitions, and editing instincts.
Built from a real voice extraction methodology refined over 15+ years of brand strategy work.
Routing
✅ Use This Skill When:
Someone wants to capture their authentic writing voice
Content "sounds too AI" and needs to match the creator's real style
Building a voice guide for consistent content across platforms
Ghostwriting or creating content on behalf of someone
User says: "write like me," "capture my voice," "this doesn't sound like me," "make it sound like me," "voice guide," "brand voice"
❌ Do NOT Use When:
Writing generic content with no voice reference → Use standard content skills
Editing existing copy for grammar/clarity only → Use copy-editing tools
Creating content strategy (what to write, not how) → Use content strategy skills
관련 스킬
Need a brand messaging framework → Use positioning skills
Better: 5-10 samples across different contexts (professional, casual, teaching, selling)
Best: Samples + a 5-minute conversation about how they think about communication
Outputs Produced:
Complete Voice DNA Profile (reusable across all future content)
Content generated in their authenticated voice
Platform-specific voice adaptations (LinkedIn vs X vs email vs proposals)
Voice consistency checklist for self-editing
Phase 1: Voice Extraction
Step 1: Collect Samples
Ask for 3-10 writing samples. The more variety, the better the profile.
Good samples:
Social posts they're proud of
Emails they wrote quickly (less filtered = more authentic)
Blog posts or articles
Podcast/video transcripts (spoken voice often reveals real patterns)
Texts or casual messages (if they're comfortable sharing)
What to tell the user:
"Send me 3-5 pieces of writing you've done. Mix of professional and casual is ideal. The ones you wrote fast without overthinking are often the most useful — that's where your real voice lives."
Step 2: Analyze Voice DNA
Read all samples and extract these 8 dimensions:
1. Sentence Architecture
Average sentence length (short and punchy? Long and flowing?)
Do they use fragments? ("Not a chance." / "Game over.")
Sentence variety pattern (short-short-long? Building momentum?)
Paragraph length preference
2. Opening Patterns (How They Start)
Do they hook with a question? A bold statement? A story? A contrarian take?
First-line energy level (explosive vs. measured)
Do they set context first or dive straight in?
3. Transition Signatures
Map their recurring bridge phrases. Everyone has them. Examples:
"Here's the thing..."
"What that means is..."
"The reality is..."
"But here's what's interesting..."
"Let me break this down..."
"So here's what happened..."
Extract at least 8-10 transition phrases from their samples. These are fingerprints.
4. Energy Mapping
Baseline energy level (calm authority? Electric enthusiasm? Quiet confidence?)
What triggers their high-energy mode?
How do they express excitement? (Exclamation marks? ALL CAPS? Power words?)
Do they use humor? What kind? (Self-deprecating? Observational? Sarcastic?)
5. Authority Zones vs. Learning Zones
This is critical and most voice tools miss it entirely.
Authority zones = Topics where they write with full confidence
Definitive language: "Here's what works," "The data shows," "What I've learned"
No hedging, no "I think maybe"
Learning zones = Topics where they're exploring
Exploratory language: "What I'm seeing," "In my experience so far," "What I'm learning"
Still confident, but framed as ongoing discovery
Map which topics fall into which zone. This prevents the AI from writing with false authority on topics the person is still learning about.
6. Vocabulary Fingerprint
Words they use often (favorites)
Words they NEVER use (allergies)
Industry jargon: do they embrace it or avoid it?
Formality level (contractions? Slang? Academic?)
Profanity comfort level
7. Structural Preferences
Do they use lists? Numbered or bulleted?
Headers or flowing prose?
Short paragraphs (1-2 sentences) or longer blocks?
Do they use bold/italic for emphasis?
Do they end with a CTA, a question, or a statement?
8. Editing Instincts
Do they tend to cut shorter or add more?
What do they delete first? (Usually: hedging words, jargon, or filler)
What's their "red flag" — the thing that makes them cringe in writing?
Phase 2: Build the Voice DNA Profile
After extraction, generate a structured Voice DNA document. This becomes the permanent reference for all future content.
Voice DNA Profile Template
# [Name]'s Voice DNA Profile
*Generated from [X] writing samples on [date]*
---
## Voice Foundation
**Core Identity:** [One sentence — who they are as a communicator]
**Natural Role:** [How they relate to their audience — teacher? Coach? Peer? Provocateur?]
**Authority Zones:** [Topics where they write with full confidence]
**Learning Zones:** [Topics where they're exploring/experimenting]
**Writing Philosophy:** [Their implicit belief about communication — extracted, not asked]
---
## Sentence Architecture
- Average sentence length: [short/medium/long]
- Uses fragments: [yes/no — with examples]
- Typical paragraph length: [1-2 / 3-4 / 5+ sentences]
- Rhythm pattern: [describe their cadence]
## Opening Patterns
- Primary hook style: [question / bold statement / story / contrarian]
- First-line energy: [1-10 scale]
- Context-setting: [dives in / sets scene first]
**Their best opening lines (from samples):**
1. "[example]"
2. "[example]"
3. "[example]"
## Transition Signatures
[List 8-12 of their actual transition phrases, organized by type]
### Authority Transitions:
- "[phrase]"
- "[phrase]"
### Energy Transitions:
- "[phrase]"
- "[phrase]"
### Story Bridges:
- "[phrase]"
- "[phrase]"
## Energy Profile
- Baseline energy: [calm / warm / enthusiastic / electric]
- High-energy triggers: [what topics fire them up]
- Excitement markers: [how they show it — exclamation marks, caps, power words]
- Humor style: [type and frequency]
## Confidence Calibration
### Write with FULL AUTHORITY when discussing:
- [topic 1]
- [topic 2]
- [topic 3]
**Voice:** Confident, definitive
**Phrases:** "[their authority phrases]"
### Write with INFORMED PERSPECTIVE when discussing:
- [topic 1]
- [topic 2]
**Voice:** Curious, exploratory but still confident
**Phrases:** "[their learning phrases]"
## Vocabulary
**Favorites:** [words they use often]
**Allergies:** [words they never use or hate]
**Jargon stance:** [embraces / avoids / selective]
**Formality:** [scale 1-10]
**Profanity:** [none / occasional / frequent]
## Structural Preferences
- Lists: [yes/no, numbered/bulleted]
- Headers: [yes/no]
- Paragraph style: [short punchy / mixed / long form]
- Emphasis: [bold / italic / caps / none]
- Endings: [CTA / question / statement / callback to opening]
## Editing Instincts
- Default edit direction: [cuts shorter / adds more]
- First things they'd delete: [hedging / jargon / filler / examples]
- Red flags: [what makes them cringe]
---
## Voice Check Questions
Before publishing as [Name], ask:
1. Energy Test: Does this feel like [their baseline] or flat?
2. Authority Test: Am I writing from confidence where they'd be confident?
3. Simplicity Test: Would [their target audience] get this immediately?
4. Landing Test: Did I land the plane or keep circling?
5. Authenticity Test: Does this sound like [Name] or like "AI writing"?
## Example Transformations
**Generic AI version:**
"[example of how AI would write it]"
**In [Name]'s voice:**
"[example rewritten in their actual voice]"
Phase 3: Generate Voice-Matched Content
With the Voice DNA Profile built, use it to generate any content type.
Content Generation Process
Load the Voice DNA Profile (read the profile before writing anything)
Identify the content type (social post, article, email, proposal, etc.)
Check confidence calibration — Is this topic in their authority zone or learning zone?
Write the first draft using their patterns:
Open with their preferred hook style
Use their transition signatures (not generic ones)
Match their sentence architecture
Apply their energy level
End with their preferred closing style
Run the Voice Check — Ask all 5 questions from the profile
Apply their editing instincts — Would they cut this shorter? Remove the hedging? Add more energy?
Platform-Specific Adaptations
The same voice adapts differently per platform. Apply these modifications ON TOP of the base voice:
LinkedIn
Slightly more structured (headers, line breaks)
Authority dialed up 10%
Hook must work in first 2 lines (before "see more")
Professional energy, not casual
End with engagement driver (question or bold statement)
X/Twitter
Punchiest version of their voice
Fragments encouraged
Energy at maximum
No hedging at all — every character counts
Thread format: each tweet must stand alone AND build
Email
Most conversational version
Can be slightly longer
Personal touches (references to shared context)
Clear CTA at the end
Warmth > authority
Long-form (Blog/Article)
Full voice expression
Stories and examples get more room
Structural preferences fully applied
Mix of authority and learning zones
Land the plane clearly at the end
Proposals/Professional Documents
Authority mode by default
Concise, confident, no filler
Proof and specifics over claims
Clear structure (they're scanning, not reading)
Phase 4: Voice Consistency Maintenance
Ongoing Calibration
When the user edits your output, note WHAT they changed. Those edits are voice data.
If they say "this doesn't sound like me," ask which specific parts feel off.
Update the Voice DNA Profile quarterly with new samples and corrections.
Common Failure Modes and Fixes
Problem
Cause
Fix
"Sounds too formal"
Formality level too high
Add more contractions, fragments, casual transitions
"Sounds too casual"
Energy overdone
Pull back excitement markers, add more structure
"Sounds like AI"
Generic transitions, no voice fingerprints
Replace ALL generic phrases with their actual transitions
"Too hedgy"
Writing in authority zone with learning-zone voice
Check confidence calibration, remove hedging language
"Not enough energy"
Baseline energy too low
Add their power words, shorten sentences, punch up hooks
"Doesn't land the plane"
Missing their closing pattern
Apply their specific ending style from the profile
Guardrails
Never fabricate voice samples. Only extract from content the user provides.
Never assume authority zones. Ask or infer from samples — don't guess.
Always produce the Voice DNA Profile first before generating content. Skip this step and the output will be generic.
If fewer than 3 samples provided, flag that the profile will be less accurate and ask for more.
Log all voice profile updates so changes can be reviewed and reverted.
The Voice DNA Profile is the user's asset. Output it in full so they own it and can use it anywhere.
Quick Start
Minimum viable run:
User provides 3 writing samples
Skill extracts Voice DNA → generates profile
User reviews profile, corrects anything off
Skill generates requested content using the profile
Time: 15-20 minutes for profile. 2-5 minutes per content piece after that.
The profile is reusable forever. Build once, use for every piece of content going forward. Update when their voice evolves.