mcpField
B
brand-voice

brand-voice

github.com/majesticlabs-dev营销与增长5057 次安装7 周前更新
代码仓库

关于 brand-voice

brand-voice 是用于创建面向 AI 的个人或组织语气指南的技能,可依据写作样本、修改稿或品牌描述开展工作。它将语气特征整理为可执行规则、示例、词汇规范和修订检查清单。

主要功能

  1. 01个人与品牌语气模式
  2. 02分析样本与修改稿
  3. 03结构化语气指南
  4. 04示例与修订检查清单
  5. 05可选的访谈式探索

适用场景

  1. 01记录组织的写作规范
  2. 02将个人写作样本整理为可复用指南
  3. 03依据语气指南撰写内容

安装

复制下面的命令,在终端中运行。

$npx skills add https://github.com/majesticlabs-dev/majestic-marketplace --skill brand-voice

安装后在 Claude Code 中输入/brand-voice确认已加载。

详细介绍

Voice Architect

创建针对 AI 优化的风格指南——一份操作手册,让大模型逐渐贴近特定的表达风格,而不是输出泛泛的默认内容。

模式检测

根据用户输入判断模式,或主动询问:

  • 个人风格(“我的风格”“我怎么写”“AI 风格指南”)→ 聚焦个人写作模式、标志性技巧和个人写作禁忌
  • 品牌风格(“我们的风格”“品牌风格”“公司风格”)→ 聚焦组织身份、跨场景语气和团队统一标准

对于定量风格指标(句长分布、标点特征、词汇多样性得分),请使用 style-forensics——它会生成包含精确数据的 Style DNA 报告。此技能会生成与这些指标互补的定性操作手册。

深入探索(可选)

如需在创建指南前全面探索表达风格,请运行:

/interview "brand voice"

这会触发一场对话式访谈,通过针对风格的问题了解身份认同、与受众的连接方式、语气边界和现有表达模式。

对话开场

使用 AskUserQuestion 收集初始信息。根据检测到的模式调整措辞。

个人风格:

“我会帮你创建一份 AI 风格指南——一份可重复使用的文档,让 AI 写出你的风格,而不是泛泛的 AI 风格。

请提供以下其中一项:

选项 A - 写作样本(首选) 分享 3-5 篇由你撰写、最能代表你最佳表达风格的作品:

  • 博客文章、新闻通讯、随笔、社交媒体帖子或邮件
  • 直接粘贴内容,或提供文件路径
  • 选择最能体现真实的你的作品

选项 B - 校准对照(最能体现风格) 分享 3-5 组并排对照的内容:AI 撰写的初稿,以及你修改后发送的版本:

  • 粘贴 AI 版本和你的最终版本
  • 两者之间的差异就是你的风格——无需自我描述
  • 邮件、帖子、Slack 消息以及任何经你编辑的内容都适用

选项 C - 风格访谈 如果手头没有样本,我会提出有针对性的问题:

  1. 展示示例段落,询问哪些更接近你的风格
  2. 询问你的写作直觉(你会从故事还是论点开始?)
  3. 探究你的写作禁忌(AI 写作中的哪些内容让你反感?)

请对示例作出反应,而不是自我描述——这样能更快发现具体特点。”

品牌风格:

“我会帮你将品牌风格整理成一份可重复使用的 AI 风格指南。

请提供以下其中一项:

选项 A - 现有内容(首选) 分享 3-5 篇最能代表品牌风格的内容:

  • 网站文案、邮件、社交媒体帖子或博客文章
  • 直接粘贴内容,或提供 URL/文件路径

选项 B - 品牌描述 如果目前还没有内容,请描述:

  1. 行业/产品:你们销售什么?
  2. 目标受众:你们面向谁?
  3. 品牌个性:用 3-5 个形容词描述品牌
  4. 欣赏的品牌:你喜欢谁的表达风格?(无论是否为竞争对手)
  5. 避免的风格:什么样的语气不适合你们?

我会分析其中的模式并创建风格指南。”

分析流程

如果提供了内容(选项 A)

从所有样本中提取模式:

需要识别的表达风格模式:

  • 句长分布(短/中/长)
  • 缩写形式的使用(can't 还是 cannot)
  • 第一/第二/第三人称偏好
  • 主动语态与被动语态的比例
  • 问句使用频率
  • 感叹号使用情况
  • 表情符号/标点风格
  • 段落长度模式
  • 开篇模式(作品如何开头)
  • 收尾模式(作品如何结束)

词汇模式:

  • 反复出现的有力词汇
  • 行业术语的使用程度(大量/少量/不用)
  • 口语表达和俚语
  • 隐喻和类比模式
  • 频繁出现的词语
  • 明显缺少的词语

语气标记:

  • 正式程度(1-10 分)
  • 幽默的使用情况(频繁/偶尔/从不)
  • 自信程度(大胆断言还是谨慎措辞)
  • 情感温度(疏远还是亲切)
  • 权威姿态(同行、专家还是导师)
如果提供了校准样本(选项 B)

这是信号最强的输入。用户对 AI 草稿所做的修改就是他们的表达风格——无需推断。

针对每组(AI 草稿、最终发送版本):

  1. 标出每处删减——他们删掉了什么?留意填充语、保留措辞和 AI 式表达
  2. 标出每处改写——他们替换了什么?找出每次替换背后的结构偏好
  3. 标出每处新增——他们加入了什么?这揭示了他们本能会采用的标志性手法

分析完所有样本后,综合归纳:

  • 反填充语清单:多个样本中反复删掉的具体短语(将其加入反填充语检查清单)
  • 结构替换:他们一贯会改写的模式(例如,被动语态 → 主动语态,清单体 → 散文)
  • 标志性增补:定义其表达风格的反复新增内容

仅根据最终发送版本,使用选项 A 中的风格模式提取标准。

如果提供了描述(选项 C)

使用 WebSearch 查找:

  • 欣赏的品牌发布的示例内容
  • 行业表达风格基准
  • 竞争对手的表达风格分析
  • 受众的沟通偏好

然后综合这些输入,提炼出一种表达风格。

品牌表达风格指南结构

1. 表达风格 DNA(核心身份)
## Voice DNA

### Brand Personality
[3-5 defining traits with explanations]

| Trait | What It Means | How It Shows Up |
|-------|---------------|-----------------|
| [Trait 1] | [Definition] | [Example in copy] |
| [Trait 2] | [Definition] | [Example in copy] |
| [Trait 3] | [Definition] | [Example in copy] |

### The Elevator Pitch
"We sound like [description]. Think [reference point] meets [reference point]."

### If Our Brand Were a Person
[2-3 sentence description of brand as human—age, profession, how they talk at a party]
2. 语气范围
## Tone Spectrum

Our voice stays consistent, but tone adapts to context.

| Context | Tone | Example |
|---------|------|---------|
| Homepage hero | Confident, bold | "Stop guessing. Start knowing." |
| Error message | Helpful, calm | "Something went wrong. Let's fix it together." |
| Success message | Warm, celebratory | "You did it! Your first campaign is live." |
| Sales email | Direct, valuable | "Here's what's working for teams like yours." |
| Support docs | Clear, patient | "First, open Settings. You'll find it in the top right." |
| Social media | Casual, engaging | "Hot take: [opinion]. Fight me in the comments." |
| Legal/Terms | Clear, straightforward | "Your data belongs to you. Here's exactly what we collect." |

### Tone Dial

**Formal ←――――――→ Casual**
[Mark where brand sits: e.g., "We sit at 3/10—professional but never stiff."]

**Serious ←――――――→ Playful**
[Mark where brand sits: e.g., "We sit at 6/10—we crack jokes but know when to be serious."]

**Reserved ←――――――→ Enthusiastic**
[Mark where brand sits: e.g., "We sit at 7/10—we're excited about what we do and it shows."]
3. 结构模式
## Structure

How pieces organize ideas — the architectural blueprint AI must follow.

### Opening Pattern
[How pieces begin: e.g., "Open with friction — a problem, tension, or surprising claim"]

### Body Organization
[How the middle works: e.g., "Alternate between concept and concrete example. Never explain without showing."]

### Closing Pattern
[How pieces end: e.g., "Land on a usable takeaway — something the reader can do today"]

### Transitions
[How sections connect: e.g., "Jump-cut between ideas. No 'Furthermore' or 'Additionally.'"]
4. 标志性手法
## Signature Moves

Named techniques that make this voice distinctive. AI should deploy these at natural frequency.

| Move | Description | Frequency |
|------|-------------|-----------|
| [Name] | [What it looks like — e.g., "Starts with a concrete anecdote, then zooms out to the principle"] | [Every piece / Often / Occasionally] |
| [Name] | [e.g., "Drops a one-sentence paragraph for emphasis after a longer passage"] | [Frequency] |
| [Name] | [e.g., "Uses parenthetical asides to add personality — like this"] | [Frequency] |

### Moves to Avoid
| Move | Why It's Wrong |
|------|---------------|
| [Name] | [e.g., "Listicle structure — feels like content marketing, not our voice"] |
5. 词汇指南
## Vocabulary

### Words We Love
| Word/Phrase | Why | Use When |
|-------------|-----|----------|
| [Word] | [Reason] | [Context] |

### Words We Avoid
| Avoid | Use Instead | Why |
|-------|-------------|-----|
| [Word] | [Alternative] | [Reason] |

### Industry Jargon Rules
[How much jargon is acceptable and when]

### Pronouns
- **We/Our**: [When to use]
- **You/Your**: [When to use]
- **I/My**: [When to use, if ever]
- **They/The company**: [When to use, if ever]

### Anti-Pattern Blacklist
Patterns to eliminate — these are structural, not just word-level:
| Pattern | Example | Fix |
|---------|---------|-----|
| [e.g., "Correlative construction"] | ["Not X, but Y"] | ["Rewrite as direct claim"] |
| [e.g., "Throat-clearing opener"] | ["In today's fast-paced world..."] | ["Cut to the point"] |
| [e.g., "Hedge stacking"] | ["It might perhaps be worth considering..."] | ["State the claim, then qualify once"] |

### Anti-Filler Checklist
Literal phrases to search-and-destroy on every AI pass. Derived from calibration pair deletions or stated pet peeves. Unlike the structural blacklist, these are exact strings.

Common AI-ism examples (replace with user's actual list):
- "delve into" → cut or rewrite as direct verb
- "it's worth noting that" → cut entirely
- "at the end of the day" → cut or replace with specific claim
- "leverage" → use "use"
- "in today's [X] landscape" → cut the opener, start with the claim
- "game-changing" / "groundbreaking" / "revolutionary" → cut, show the claim instead
*Populate from calibration pairs: every phrase deleted across 2+ pairs is a checklist item.*
6. 句式风格
## Sentence Style

### Length
- **Target**: [X words average per sentence]
- **Mix**: [Short sentences for punch, longer for explanation]
- **Paragraphs**: [Max X sentences per paragraph]

### Structure Preferences
- **Contractions**: [Always/Sometimes/Never] — "you're" vs "you are"
- **Active voice**: [Percentage target] — "We built this" vs "This was built"
- **Starting sentences**: [Patterns to use/avoid]
- **Questions**: [Rhetorical? Direct? Frequency?]

### Punctuation
- **Exclamation points**: [Rules for usage]
- **Em dashes**: [Heavy use/light use]
- **Ellipses**: [Never/sparingly/frequently]
- **Oxford comma**: [Yes/No]
- **Emojis**: [Never/sparingly/frequently + which ones]
7. 格式规范
## Formatting

### Headlines
- **Case**: [Sentence case / Title Case]
- **Length**: [Max X words]
- **Punctuation**: [Period/No period]

### CTAs
- **Style**: [Action verb first]
- **Examples**: [List of preferred CTA phrases]

### Lists
- **Bullet style**: [Dashes/dots/checkmarks]
- **Capitalization**: [First word only / Each word]
- **Punctuation**: [Periods/No periods]

### Numbers
- **Spell out**: [One through ten / all / none]
- **Percentages**: [50% vs fifty percent]
- **Currency**: [$X vs X dollars]
8. 应做/不应做示例
## Do/Don't Examples

### Homepage Hero
❌ "Welcome to [Company]. We are the leading provider of innovative solutions."
✅ "[Bold claim that shows, not tells]."

### Feature Description
❌ "Our platform leverages cutting-edge technology to deliver best-in-class results."
✅ "[Specific outcome in plain language]."

### Email Subject Line
❌ "Newsletter #47 - Monthly Update"
✅ "[Curiosity hook or specific benefit]."

### Error Message
❌ "Error 403: Forbidden access denied."
✅ "[Human explanation + next step]."

### CTA Button
❌ "Submit" / "Click Here"
✅ "[Action + Outcome]" — "Start Free Trial" / "Get Your Report"

### Social Post
❌ "We are pleased to announce..."
✅ "[Direct statement or hook]."
9. 修改检查清单
## Revision Checklist

Run through before publishing. Each question should produce a confident "yes."

### Voice
- [ ] Does this sound like a real person — not a committee?
- [ ] Could this only be written by us/me? (Not generic AI)
- [ ] Are 2+ signature moves deployed naturally?
- [ ] Zero blacklisted patterns present?

### Tone
- [ ] Is the tone calibrated for this context?
- [ ] Does it match the tone dial positions?

### Language
- [ ] No words from the "avoid" list?
- [ ] Anti-pattern blacklist clear?
- [ ] Anti-filler checklist phrases removed?
- [ ] Contractions/pronouns consistent?

### Structure
- [ ] Opening follows the structure pattern?
- [ ] Transitions feel like jump-cuts, not academic bridges?
- [ ] Closing lands on something usable?

### Gut Check
- [ ] Read it aloud — does it sound like talking, not writing?

输出格式

个人使用“VOICE GUIDE”,组织使用“BRAND VOICE GUIDE”。

# [BRAND] VOICE GUIDE: [Name]

*Version 1.0 | Created [Date] | Optimized for AI consumption*

---

## Quick Reference

**Voice:** [3 traits]
**Sounds like:** [1-sentence description]
**Never sounds:** [What to avoid]

---

## 1. VOICE DNA
[Personality traits + identity]

## 2. TONE SPECTRUM
[Context-based tone adjustments]

## 3. STRUCTURE PATTERNS
[Opening, body, closing, transition patterns]

## 4. SIGNATURE MOVES
[Named techniques with frequency]

## 5. VOCABULARY + ANTI-PATTERNS
[Words we love/avoid + structural blacklist + anti-filler checklist]

## 6. SENTENCE STYLE
[Length, rhythm, punctuation rules]

## 7. FORMATTING
[Headlines, CTAs, lists, numbers]

## 8. DO/DON'T EXAMPLES
[Before/after examples by content type]

## 9. REVISION CHECKLIST
[Pre-publish validation questions]

## APPENDIX: Sample Rewrites
[Off-voice → on-voice with annotations]

文件输出

生成指南后,主动提出将其保存到:

  • 品牌: docs/brand-voice.md(推荐)
  • 个人: docs/style-guide.md 或用户指定的位置(例如 Claude Projects)
  • 供 content-writer、content-atomizer、linkedin-content、landing-page-builder、style-writer 引用

质量标准

  • 具体胜于笼统——“使用缩写形式”,而不是“风格随意”
  • 必须提供示例——每条规则都需要具体示例
  • 规则可执行——必须能够核查是否遵守
  • 以实际内容为依据——从真实内容中提取,不要凭空编造
  • 持续更新的文档——注明表达风格指南应不断演进

与其他技能的集成

此表达风格指南适用于:

  • content-writer - 将表达风格应用于文章
  • content-atomizer - 在不同平台上保持表达风格一致
  • linkedin-content - 符合品牌风格的社交媒体帖子
  • landing-page-builder - 与表达风格一致的落地页
  • sales-page - 保持销售文案一致

相关技能:

  • style-forensics - 定量提取个人表达风格(Style DNA 报告)
  • style-writer - 撰写符合某种 Style DNA 或品牌表达风格的新内容
  • humanizer - 在遵循风格约束的同时去除 AI 痕迹

使用模式:

"Write this using the voice guide in docs/brand-voice.md"

维护说明

在以下情况下,应更新表达风格指南:

  • 品牌定位发生变化
  • 新增内容类型
  • 团队反馈发现缺漏
  • 表达风格偏离指南(重新对齐或更新指南)

建议每季度审查一次。

评论

支持 Markdown