Brand Voice

Brand Voice Skill

Purpose

Maintain consistent brand voice across all AI-generated content by:

1. Reading voice profile from BRAND_VOICE.md
2. Applying tone, vocabulary, and style rules
3. Flagging off-brand content
4. Suggesting on-brand alternatives

When to Activate

Activation conditions:

• Writing marketing copy, social posts, emails
• Creating customer-facing content
• Any branded communication
• User mentions "brand voice" or "on-brand"
• BRAND_VOICE.md exists in workspace

Setup: BRAND_VOICE.md

This skill requires a BRAND_VOICE.md file in your workspace. Create one using the template below.

Template

# Brand Voice Profile

## Brand Identity

**Brand Name:** [Your brand name]
**Tagline:** [Your tagline if any]
**Industry:** [Your industry]
**Target Audience:** [Who you're talking to]

## Voice Attributes

Rate each attribute 1-5:

| Attribute | Level (1-5) | Notes |
|-----------|-------------|-------|
| Formal ↔ Casual | | |
| Serious ↔ Playful | | |
| Reserved ↔ Enthusiastic | | |
| Technical ↔ Simple | | |
| Traditional ↔ Modern | | |

## Tone Guidelines

### We Sound Like:
- [Descriptor 1, e.g., "A knowledgeable friend"]
- [Descriptor 2, e.g., "Confident but not arrogant"]
- [Descriptor 3]

### We Don't Sound Like:
- [Anti-descriptor 1, e.g., "A corporate robot"]
- [Anti-descriptor 2, e.g., "Pushy salesperson"]
- [Anti-descriptor 3]

## Vocabulary

### Words We Use:
- [Word/phrase 1]
- [Word/phrase 2]
- [Word/phrase 3]

### Words We Avoid:
- [Word 1] → Use [alternative] instead
- [Word 2] → Use [alternative] instead
- [Word 3] → Use [alternative] instead

## Grammar & Style

- **Contractions:** [Yes/No/Sometimes]
- **Exclamation marks:** [Rarely/Sometimes/Often]
- **Emoji:** [Never/Sparingly/Freely]
- **Sentence length:** [Short/Medium/Varied]
- **Oxford comma:** [Yes/No]

## Sample Phrases

### On-Brand Examples:
> "[Example sentence that nails your voice]"

> "[Another example]"

### Off-Brand Examples:
> "[Example of what NOT to write]"
❌ Problem: [Why this is off-brand]

## Channel Variations

### Email
- [Any email-specific rules]

### Social Media
- [Platform-specific rules]

### Website
- [Website copy rules]

## Reference Content

[Link or paste 2-3 examples of your best on-brand content]

Voice Enforcement Process

Step 1: Load Profile

When generating content, first read BRAND_VOICE.md to understand:

• Voice attributes (formal/casual, serious/playful, etc.)
• Vocabulary preferences
• Style rules
• Sample phrases

Step 2: Generate Draft

Create content following the profile guidelines.

Step 3: Voice Check

Before delivering, verify:

• [ ] Tone matches voice attributes
• [ ] No words from "avoid" list
• [ ] Grammar rules followed
• [ ] Sounds like "we sound like" descriptors
• [ ] Doesn't sound like "we don't sound like" descriptors

Step 4: Flag Issues

If content might be off-brand:

⚠️ Voice Check:
- [Issue 1]: Used "[word]" — consider "[alternative]"
- [Issue 2]: Tone feels [too formal/casual] for your brand

Voice Calibration

Formal ↔ Casual Scale

| Level | Example | |-------|---------| | 1 (Very Formal) | "We appreciate your inquiry and will respond promptly." | | 2 (Professional) | "Thanks for reaching out. We'll get back to you soon." | | 3 (Friendly Professional) | "Thanks for your message! We'll be in touch shortly." | | 4 (Casual) | "Hey! Got your message — we'll hit you back soon." | | 5 (Very Casual) | "Yo! Message received 👍 Talking soon!" |

Serious ↔ Playful Scale

| Level | Example | |-------|---------| | 1 (Very Serious) | "Our solution addresses critical business challenges." | | 2 (Professional) | "Our solution helps businesses solve real problems." | | 3 (Balanced) | "We built this to solve problems you actually have." | | 4 (Light) | "We made something useful. Crazy, right?" | | 5 (Very Playful) | "We fixed the thing! 🎉 You're welcome." |

Content Templates by Channel

Email Subject Lines

[Brand voice applied]:
- Formal: "Your [Topic] Update: [Key Point]"
- Casual: "[Key Point] — here's what you need to know"
- Playful: "[Key Point] 👀 You'll want to see this"

Social Media Posts

[Brand voice applied]:
- Formal: "[Statement]. Learn more: [link]"
- Casual: "[Conversational take]. [link]"
- Playful: "[Witty observation] 🔥 [link]"

Error Messages

[Brand voice applied]:
- Formal: "An error has occurred. Please try again."
- Casual: "Oops, something went wrong. Give it another shot?"
- Playful: "Well, that didn't work 😅 Mind trying again?"

Voice Consistency Checks

Quick Check (Before Every Output)

1. Read first sentence aloud — does it sound like your brand?
2. Check for any "avoid" words
3. Verify tone matches context

Deep Check (For Important Content)

1. Compare against sample phrases
2. Read entire piece for consistency
3. Check all vocabulary
4. Verify channel-specific rules applied
5. Get second opinion if high-stakes

Handling Edge Cases

When Tone Conflicts with Message

Sometimes the message requires a different tone (e.g., playful brand delivering serious news).

Approach:

1. Maintain brand voice foundation
2. Adjust intensity appropriately
3. Don't abandon voice entirely

Example:

Off-brand (too different):
"We regret to inform you of a service disruption."

On-brand (adjusted):
"Hey — we hit a snag. [Service] is down right now. 
We're on it and will update you in 30 minutes."

When Writing for Different Audiences

Same brand, different audience = same voice, different vocabulary.

Example:

To customers: "Your dashboard just got way faster."
To developers: "Dashboard load time reduced 60% via lazy loading."

Response Principles

1. Profile first — Always check BRAND_VOICE.md before writing
2. Consistency over perfection — Slight off-brand is better than inconsistent
3. Context matters — Adjust intensity, not identity
4. Flag don't fix silently — Tell user if something's off-brand
5. Learn patterns — Note recurring issues to improve profile

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: mcltyl
• Source: mcltyl/brand-voice-skills
• License: MIT

Install and usage instructions live in the source repository linked above.