voice-apply

name	voice-apply
description	Applies a voice profile to transform content. Use when user asks to write in a specific voice, match a tone, apply a style, or transform content to sound like a particular voice profile.
version	1.0.0
platforms	["all"]

Voice Apply Skill

Purpose

Transform content to match a specified voice profile. This skill loads voice profiles and applies their characteristics (tone, vocabulary, structure, perspective) to new or existing content.

Critical rule: Every piece of output MUST pass the AI writing tells check defined in references/ai-tells.md. This is non-negotiable. If the output contains any banned vocabulary, structures, transitions, or punctuation patterns from that reference, rewrite until clean.

When This Skill Applies

• User asks to "write in X voice" or "use Y tone"
• User wants to "make this sound more [casual/formal/technical/etc.]"
• User provides content and asks to transform its style
• User references a voice profile by name
• User wants content to match a specific audience or context

Trigger Phrases

Natural Language	Action
"Write this in technical voice"	Apply technical-authority profile
"Make it more casual"	Apply casual-conversational or calibrate toward casual
"This needs to sound executive"	Apply executive-brief profile
"Explain like I'm a beginner"	Apply friendly-explainer profile
"Use the [profile-name] voice"	Load and apply named profile
"Transform this to match [example]"	Analyze example, apply derived voice

Voice Profile Locations

Skill checks these locations (in order):

1. Project: .aiwg/voices/
2. User: ~/.config/aiwg/voices/
3. Built-in: voice-framework/voices/templates/

Built-in Voice Profiles

Profile	Description	Best For
technical-authority	Direct, precise, confident	Docs, architecture, engineering
friendly-explainer	Approachable, encouraging	Tutorials, onboarding, education
executive-brief	Concise, outcome-focused	Business cases, stakeholder comms
casual-conversational	Relaxed, personal	Blog posts, social, newsletters

Application Process

1. Load Voice Profile

# Load from YAML
profile = load_voice_profile("technical-authority")

2. Analyze Source Content (if transforming)

• Current tone characteristics
• Vocabulary patterns
• Structure patterns
• Gap analysis vs target voice

3. Apply Voice Characteristics

Tone Calibration:

• Adjust formality level (word choice, contractions)
• Calibrate confidence (hedging vs assertion)
• Set warmth (clinical vs personable)
• Tune energy (measured vs enthusiastic)

Vocabulary Transformation:

• Replace words per prefer/avoid guidance
• Introduce domain terminology naturally
• Weave in signature phrases where appropriate

Structure Adjustment:

• Modify sentence length distribution
• Adjust paragraph breaks
• Add/remove lists, examples, analogies as specified

Perspective Shift:

• Adjust narrative person (I, we, you, they)
• Calibrate opinion expression
• Set reader relationship tone

4. AI Tells Elimination Pass (MANDATORY)

After applying the voice profile, run every piece of output through the full AI tells check. This is the most important step. Reference: references/ai-tells.md

Vocabulary check: Scan output for every word in the banned lists:

• Banned verbs: delve, leverage, utilize, harness, unlock, unleash, empower, facilitate, foster, bolster, optimize, streamline, navigate, spearhead, underscore, illuminate, elucidate, embark, unravel, elevate, reimagine, revolutionize, transcend, resonate, showcase, craft, embrace, thrive, unveil, champion, and all others on the list
• Banned adjectives: multifaceted, seamless, robust, comprehensive, cutting-edge, holistic, meticulous, groundbreaking, transformative, innovative, compelling, invaluable, paramount, unprecedented, impactful, game-changing, and all others on the list
• Banned nouns: tapestry, realm, testament, beacon, myriad, landscape, ecosystem, paradigm, nexus, catalyst, synergy, roadmap, cornerstone, bedrock, pinnacle, plethora, stakeholders, and all others on the list
• Banned adverbs: furthermore, moreover, additionally, notably, crucially, importantly, consequently, seamlessly, fundamentally, undoubtedly, and all others on the list

Transition check: Remove all banned transitions:

• Throat-clearing: "in today's digital age", "in a world where", "when it comes to"
• Pedagogical: "let's dive in", "let's unpack this", "great question!"
• Signposting: "it's important to note", "it's worth noting", "this underscores"
• Fake-suspense: "here's the thing", "the best part?", "here's the kicker"
• Hype: "game-changer", "pushing the boundaries", "paving the way"
• Conclusions: "in conclusion", "to wrap up", "at the end of the day"
• Bot closers: "I hope this helps!", "feel free to reach out"

Structure check: Eliminate banned patterns:

• The "It's not X, it's Y" negation pattern (the single most prominent AI tell)
• Self-posed questions answered immediately ("The best part? It's this.")
• Anaphora abuse (same opener 3+ times in a row)
• Present participle trailing clauses ("highlighting its importance", "reflecting broader trends")
• False ranges ("From innovation to implementation")
• Hedge-stacking (multiple qualifiers in one sentence)
• Consecutive adverbial sentence starters

Punctuation check:

• Remove em dashes. Replace with commas, periods, parentheses, or colons.
• Remove semicolons (unless user explicitly uses them)
• Remove Oxford commas (unless user explicitly uses them)
• Replace Unicode ellipsis with three dots or restructure

Rhythm check:

• Verify sentence length varies (mix 5-word punches with 25-word flows)
• Verify paragraph length varies (some 1-sentence, some 5-sentence)
• Break any Rule of Three that appears back-to-back
• Check for uniform subsection length and vary it

Formatting check:

• No bold-first bullet points (every bullet starting with "Word: explanation")
• No excessive headers for content that doesn't need them
• No erratic bolding of random words
• No emoji decoration in professional contexts

Tone check:

• No relentless positivity. Include criticism where warranted.
• No false balance. Make bold claims when the evidence supports them.
• No emotional flatness. Let genuine reactions show.
• No elevated register for simple ideas.

Self-reference check:

• No mentions of being an AI or language model
• No "as someone who..."
• No disclaimers about AI limitations
• No "as of my last knowledge update"
• No "Certainly, here are..."
• Use contractions naturally unless context is genuinely formal

5. Verify Authenticity Markers

Ensure output includes profile's authenticity characteristics:

• Acknowledges uncertainty (if specified)
• Shows tradeoffs (if specified)
• Uses specific numbers (if specified)
• References constraints (if specified)

6. Final Read-Aloud Test

Read the output in your head. If any sentence sounds like it could appear in any other AI-generated text, rewrite it until it sounds like a specific person said it.

Usage Examples

Apply Named Voice

User: "Write release notes in technical-authority voice"

Process:
1. Load technical-authority.yaml
2. Generate release notes with:
   - Precise technical terminology
   - Specific version numbers
   - Direct, confident statements
   - Tradeoff acknowledgments where relevant
3. AI tells elimination pass
4. Final read-aloud test

Transform Existing Content

User: "Make this documentation more friendly for beginners"

Input: "The API endpoint accepts a JSON payload containing the requisite parameters..."

Process:
1. Load friendly-explainer.yaml
2. Analyze: formal, technical, passive
3. Transform to: casual, accessible, active
4. AI tells elimination pass

Output: "To use this endpoint, send it some JSON with the info it needs..."

Calibrate Voice

User: "This is too formal, dial it back 30%"

Process:
1. Identify current formality (~0.8)
2. Calculate target (0.8 - 0.3 = 0.5)
3. Adjust vocabulary and structure for medium formality
4. AI tells elimination pass

Voice Blending

Combine multiple profiles:

User: "Write this with 70% technical-authority and 30% friendly-explainer"

Process:
1. Load both profiles
2. Weighted merge:
   - tone.formality: 0.7 * 0.7 + 0.3 * 0.3 = 0.58
   - tone.warmth: 0.7 * 0.3 + 0.3 * 0.8 = 0.45
   - etc.
3. Apply merged profile
4. AI tells elimination pass (always runs last)

Output Format

When reporting voice application:

Voice Applied: technical-authority

Transformations:
- Formality: 0.4 -> 0.7 (increased)
- Confidence: 0.5 -> 0.9 (increased)
- Vocabulary: 12 replacements
- Structure: Added 2 examples, removed 1 rhetorical question

AI Tells Check:
- Banned vocabulary removed: 3 instances
- Em dashes replaced: 2 instances
- Banned structures fixed: 1 negation pattern
- Sentence length variance: OK (std dev 7.4)
- Paragraph length variance: OK (varied)

Authenticity Check:
- Acknowledges tradeoffs
- Uses specific numbers
- References constraints

References

• AI Writing Tells (MANDATORY): references/ai-tells.md