brand-voice-extractor
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseBrand Voice Extractor
品牌调性提取工具
Analyze a company's published content to extract their brand voice and writing style. Reads their top content pieces and produces actionable guidelines for matching their voice in future content, outreach, or campaigns.
分析企业已发布的内容，提取其品牌调性与写作风格。读取其优质内容，生成可落地的规范，以便在未来的内容、开发信或营销活动中匹配该品牌调性。
Quick Start
快速开始
Extract brand voice for [company]. Use their blog at [url].Or with content already cataloged:
Extract brand voice for [client]. Use the content inventory at clients/[client]/research/content-inventory.json.Extract brand voice for [company]. Use their blog at [url].或使用已整理好的内容：
Extract brand voice for [client]. Use the content inventory at clients/[client]/research/content-inventory.json.Inputs
输入项
| Input | Required | Source |
|---|---|---|
| Content URLs | Yes | User provides, or pulled from site-content-catalog output |
| Company name | Yes | For context in the analysis |
| Number of pages | No | Default: 15. How many pages to analyze. |
| 输入项 | 是否必填 | 来源 |
|---|---|---|
| 内容URL | 是 | 用户提供，或从site-content-catalog输出中获取 |
| 企业名称 | 是 | 用于分析时提供上下文 |
| 页面数量 | 否 | 默认值：15。需分析的页面数量。 |
Process
流程
Phase 1: Select Content to Analyze
阶段1：选择待分析的内容
If content URLs are provided directly, use those. Otherwise:
- Read the content inventory from output
site-content-catalog - Select a diverse sample of 10-20 pages, prioritizing:
- Blog posts (primary voice indicator)
- Landing pages (marketing voice)
- Case studies (storytelling voice)
- Mix of recent and older content (to detect voice evolution)
- Mix of topics (to see consistency across subjects)
Selection heuristic:
- 8-10 blog posts (mix of how-to, opinion, product updates)
- 2-3 landing pages (homepage, product page, solutions page)
- 2-3 case studies or customer stories (if available)
- 1-2 comparison/vs pages (if available)
如果直接提供了内容URL，则使用这些URL。否则：
- 读取输出的内容清单
site-content-catalog - 选择10-20页多样化的样本，优先选择：
- 博客文章（品牌调性的核心体现）
- 落地页（营销调性）
- 案例研究（叙事调性）
- 近期与早期内容的混合（以检测调性的演变）
- 不同主题的混合（以查看跨主题的调性一致性）
选择规则：
- 8-10篇博客文章（涵盖操作指南、观点类、产品更新等类型）
- 2-3个落地页（首页、产品页、解决方案页）
- 2-3个案例研究或客户故事（如果有）
- 1-2个对比类页面（如果有）
Phase 2: Fetch and Extract Text
阶段2：获取并提取文本
For each selected URL:
- WebFetch the page
- Extract the main content body (strip nav, footer, sidebar)
- Store: title, URL, raw text, word count
对于每个选中的URL：
- 通过WebFetch获取页面
- 提取主要内容主体（去除导航栏、页脚、侧边栏）
- 存储：标题、URL、原始文本、字数
Phase 3: Analyze Voice Dimensions
阶段3：分析调性维度
Analyze across these dimensions:
从以下维度进行分析：
A) Tone
A) 语气
- Formality spectrum: Casual ↔ Professional ↔ Academic
- Emotional register: Excited ↔ Measured ↔ Dry
- Authority stance: Peer/friend ↔ Expert/teacher ↔ Institution
- Humor usage: Frequent ↔ Occasional ↔ None
- Directness: Direct/bold ↔ Hedged/diplomatic
- 正式程度范围： 随意 ↔ 专业 ↔ 学术
- 情感基调： 兴奋 ↔ 沉稳 ↔ 平淡
- 权威定位： 同行/朋友 ↔ 专家/导师 ↔ 机构
- 幽默使用： 频繁 ↔ 偶尔 ↔ 无
- 直接性： 直接/大胆 ↔ 委婉/圆滑
B) Vocabulary & Language
B) 词汇与语言
- Reading level: Approximate grade level (simple vs. complex)
- Jargon usage: Heavy industry jargon ↔ Plain language
- Technical depth: Assumes expertise ↔ Explains everything
- Power words: Common persuasion/action words they favor
- Banned patterns: Words or phrases they conspicuously avoid
- Unique vocabulary: Distinctive terms or phrases they use repeatedly
- 阅读难度： 近似年级水平（简单 vs 复杂）
- 行话使用： 大量行业行话 ↔ 平实语言
- 技术深度： 假设读者具备专业知识 ↔ 详细解释所有内容
- 强力词汇： 他们偏好的常用说服/行动词汇
- 禁用模式： 他们明显避免使用的词汇或短语
- 独特词汇： 他们重复使用的特有术语或短语
C) Sentence Structure
C) 句子结构
- Average sentence length: Short/punchy ↔ Long/complex
- Paragraph length: 1-2 sentences ↔ 3-4 ↔ 5+
- Opening patterns: How they start articles (question, stat, story, bold claim)
- Transition style: How they connect ideas
- Use of fragments: Do they use incomplete sentences for emphasis?
- 平均句子长度： 简短有力 ↔ 冗长复杂
- 段落长度： 1-2句 ↔ 3-4句 ↔ 5句以上
- 开篇模式： 文章的开头方式（问题、数据、故事、大胆断言）
- 过渡风格： 他们连接观点的方式
- 碎片句使用： 他们是否使用不完整句子来强调内容？
D) Formatting Patterns
D) 格式模式
- Headers: Frequency, style (question-based, how-to, numbered)
- Lists: Bullets vs. numbered, frequency
- Bold/italic: How they use emphasis
- Images/media: Frequency, types (screenshots, illustrations, photos)
- CTAs: Placement, style, frequency, language used
- Pull quotes/callouts: Do they use them?
- 标题： 频率、风格（问题式、操作指南式、编号式）
- 列表： 项目符号 vs 编号，使用频率
- 粗体/斜体： 他们如何使用强调格式
- 图片/媒体： 频率、类型（截图、插画、照片）
- CTA： 位置、风格、频率、使用的语言
- 引用/标注： 他们是否使用这些格式？
E) Content Structure
E) 内容结构
- Typical article length: Short (<800), Medium (800-1500), Long (1500+)
- Introduction style: Hook type, length
- Conclusion style: Summary, CTA, open question
- Use of data/stats: Frequent ↔ Rare
- Use of examples: Frequent ↔ Rare
- Storytelling: Narrative-driven ↔ Information-driven
- 典型文章长度： 短篇（<800字）、中篇（800-1500字）、长篇（1500字以上）
- 引言风格： 钩子类型、长度
- 结论风格： 总结、CTA、开放式问题
- 数据/统计使用： 频繁 ↔ 罕见
- 示例使用： 频繁 ↔ 罕见
- 叙事性： 叙事驱动 ↔ 信息驱动
F) Persona & Audience
F) 用户画像与受众
- Who they write for: Inferred target reader (role, seniority, industry)
- Assumed knowledge level: Beginner ↔ Intermediate ↔ Expert
- Point of view: First person singular (I) ↔ First person plural (we) ↔ Second person (you) ↔ Third person
- Reader relationship: Peer ↔ Teacher ↔ Service provider
- 目标读者： 推断出的目标读者（角色、职级、行业）
- 假设知识水平： 初学者 ↔ 中级 ↔ 专家
- 视角： 第一人称单数（我） ↔ 第一人称复数（我们） ↔ 第二人称（你） ↔ 第三人称
- 读者关系定位： 同行 ↔ 导师 ↔ 服务提供者
Phase 4: Generate Brand Voice Profile
阶段4：生成品牌调性档案
Produce a Markdown document with this structure:
markdown
undefined生成如下结构的Markdown文档：
markdown
undefinedBrand Voice Profile: [Company Name]
Brand Voice Profile: [Company Name]
Analyzed: [Date] | Content pieces analyzed: [N]
Sources: [list of URLs analyzed]
Analyzed: [Date] | Content pieces analyzed: [N]
Sources: [list of URLs analyzed]
Voice Summary (2-3 sentences)
Voice Summary (2-3 sentences)
[Company] writes in a [tone] voice that [description]. Their content targets
[audience] and assumes [knowledge level]. The overall feel is [adjectives].
[Company] writes in a [tone] voice that [description]. Their content targets
[audience] and assumes [knowledge level]. The overall feel is [adjectives].
Tone Profile
Tone Profile
| Dimension | Position | Evidence |
|---|---|---|
| Formality | [e.g., Professional-casual] | [Example quote] |
| Emotional Register | [e.g., Measured, occasionally excited] | [Example] |
| Authority | [e.g., Expert/teacher] | [Example] |
| Humor | [e.g., Rare, dry when used] | [Example] |
| Directness | [e.g., Very direct, bold claims] | [Example] |
| Dimension | Position | Evidence |
|---|---|---|
| Formality | [e.g., Professional-casual] | [Example quote] |
| Emotional Register | [e.g., Measured, occasionally excited] | [Example] |
| Authority | [e.g., Expert/teacher] | [Example] |
| Humor | [e.g., Rare, dry when used] | [Example] |
| Directness | [e.g., Very direct, bold claims] | [Example] |
Language & Vocabulary
Language & Vocabulary
Reading Level
Reading Level
[Grade level estimate and what that means]
[Grade level estimate and what that means]
Signature Phrases
Signature Phrases
- "[phrase 1]" — used frequently to [purpose]
- "[phrase 2]" — recurring pattern in [context]
- "[phrase 1]" — used frequently to [purpose]
- "[phrase 2]" — recurring pattern in [context]
Jargon & Technical Depth
Jargon & Technical Depth
[How much industry jargon they use, how they handle technical concepts]
[How much industry jargon they use, how they handle technical concepts]
Words They Love
Words They Love
[List of frequently used power words, adjectives, verbs]
[List of frequently used power words, adjectives, verbs]
Words They Avoid
Words They Avoid
[Notable absences or patterns they steer away from]
[Notable absences or patterns they steer away from]
Structure & Formatting
Structure & Formatting
Typical Article Structure
Typical Article Structure
[Outline of how their articles are typically organized]
[Outline of how their articles are typically organized]
Sentence & Paragraph Style
Sentence & Paragraph Style
- Average sentence length: [X words]
- Typical paragraph: [X sentences]
- Notable patterns: [fragments, rhetorical questions, etc.]
- Average sentence length: [X words]
- Typical paragraph: [X sentences]
- Notable patterns: [fragments, rhetorical questions, etc.]
Formatting Habits
Formatting Habits
- Headers: [style]
- Lists: [frequency and style]
- Emphasis: [bold/italic patterns]
- CTAs: [where, how often, what language]
- Headers: [style]
- Lists: [frequency and style]
- Emphasis: [bold/italic patterns]
- CTAs: [where, how often, what language]
Audience & Persona
Audience & Persona
Target Reader
Target Reader
[Role, seniority, industry, pain points they address]
[Role, seniority, industry, pain points they address]
Knowledge Assumptions
Knowledge Assumptions
[What they assume the reader already knows]
[What they assume the reader already knows]
Point of View
Point of View
[I/we/you usage and what it signals]
[I/we/you usage and what it signals]
Writing Guidelines (Actionable)
Writing Guidelines (Actionable)
Use these guidelines when writing content, outreach, or campaigns for [Company]:
Use these guidelines when writing content, outreach, or campaigns for [Company]:
Do
Do
- [Guideline 1 with example]
- [Guideline 2 with example]
- [Guideline 3 with example]
- [Guideline 1 with example]
- [Guideline 2 with example]
- [Guideline 3 with example]
Don't
Don't
- [Anti-pattern 1]
- [Anti-pattern 2]
- [Anti-pattern 3]
- [Anti-pattern 1]
- [Anti-pattern 2]
- [Anti-pattern 3]
Voice Samples
Voice Samples
Their style:
[2-3 representative quotes from their content that exemplify the voice]
How to match it:
[2-3 example sentences written in their voice about a neutral topic]
undefinedTheir style:
[2-3 representative quotes from their content that exemplify the voice]
How to match it:
[2-3 example sentences written in their voice about a neutral topic]
undefinedTips
提示
- 15 pages is the sweet spot. Fewer than 10 won't capture enough variation. More than 25 adds cost without much signal.
- Blog posts are the best voice signal. Landing pages are more formulaic. Blog posts show the authentic voice.
- Look for consistency AND inconsistency. If their tone shifts dramatically between content types, note it — they may have multiple voice modes.
- Check for ghost-written content. If some posts feel dramatically different, they may use external writers. Flag this in the analysis.
- This skill has no code script. It's an agent-executed skill — the AI agent reads the content via WebFetch and performs the analysis directly. The structured output template above guides the analysis.
- 15页是最优数量。少于10页无法捕捉足够的调性变化，超过25页会增加成本但不会显著提升分析信号。
- 博客文章是最佳调性信号。落地页更具公式化，博客文章能体现真实的品牌调性。
- 寻找一致性与不一致性。如果不同类型内容的调性差异巨大，请记录下来——他们可能有多种调性模式。
- 检查是否为代笔内容。如果部分文章风格差异明显，可能使用了外部作者，请在分析中标记这一点。
- 本技能无代码脚本。它是由Agent执行的技能——AI Agent通过WebFetch读取内容并直接进行分析。上述结构化输出模板用于指导分析过程。
Dependencies
依赖项
- Web fetch capability (for reading content pages)
- Optional: output (for selecting which content to analyze)
site-content-catalog - No API keys or paid tools required
- WebFetch功能（用于读取内容页面）
- 可选：输出（用于选择待分析的内容）
site-content-catalog - 无需API密钥或付费工具