Token导航 LogoToken导航

Agent Skill · 研究检索

voice-matching-wizard(语音匹配向导)

用于辅助音频、音乐、语音转写、语音合成或声音素材处理。它适合让 Agent 生成配乐说明、整理音频流程、调用语音工具或处理播客和视频配音素材。使用时需要确认输入音频来源、输出格式、时长和模型限制;涉及人声克隆、版权音乐或公开发布时,应先核对授权和合规边界。

voice-matching-wizard 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为权限未标注;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 59,221。

总安装 392周安装 16GitHub Stars 9下载 125核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:voice-matching-wizard(语音匹配向导)
来源仓库:https://github.com/cdeistopened/skill-stack
仓库路径:skills/voice-matching-wizard
安装命令:
npx skills add https://github.com/cdeistopened/skill-stack --skill voice-matching-wizard
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/cdeistopened/skill-stack --skill voice-matching-wizard

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

创建可重用的语音匹配技能,捕捉任何写作风格的识别特征。适用于品牌声音建立、作者风格复制和内容一致性场景。使用时需收集样本、提取多维模式和生成可部署的技能文件。通过 GitHub 安装,支持主流宿主环境,建议核对样本授权和用途合规性。

voice-matching-wizard 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/voice-matching-wizard

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/cdeistopened/skill-stack --skill voice-matchin…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    权限没有明确标注,第一次使用时只给一个无关紧要的任务,观察它到底读了哪些文件和网络。

常见问题

6 条

voice-matching-wizard 怎么安装?

在终端执行 npx skills add https://github.com/cdeistopened/skill-stack --skill voice-matching-wizard。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

voice-matching-wizard 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

voice-matching-wizard 需要什么权限?

权限级别没有明确标注。第一次使用时建议只给无关紧要的任务,观察它实际读写了哪些内容。

voice-matching-wizard 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

voice-matching-wizard 用的人多吗?

公开数据是累计安装 392 次,近一周安装 16 次,下载 125 次,GitHub 9 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

voice-matching-wizard 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
voice-matching-wizard

语音匹配向导 · 本页

3929权限未标注未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Voice Matching Wizard

Create a voice skill that captures the patterns, rhythms, and sensibilities of any writing style.

What This Does

Voice matching is the art of identifying what makes writing *recognizable*. Not just the obvious markers—vocabulary, sentence length—but the deeper patterns: how ideas unfold, where the writer pauses, what they leave unsaid.

This wizard walks you through the process of:

  1. Gathering representative samples
  2. Extracting patterns across multiple dimensions
  3. Synthesizing a usable voice profile
  4. Generating a working skill file

The output is a voice-[name].skill.md you can use alongside other skills (like anti-ai-writing) for consistent, authentic content.


Before You Begin

What makes a good sample:

  • At least 500 words
  • Writing you're proud of (or writing you want to emulate)
  • Representative of the voice you want to capture
  • From the same medium (all newsletters, all blog posts, etc.)

What to avoid:

  • Heavily edited committee writing
  • Content written under constraints
  • Mixed formats (tweets + longform together)
  • Very old work that doesn't reflect current voice

How many samples:

  • 2-3 samples: Basic patterns
  • 4-5 samples: Good coverage
  • 10+: Comprehensive capture

Three Paths Through This Wizard

Path A: "I have samples of my own writing"

You want to codify your existing voice so AI can match it consistently. → Skip to Phase 1: Gather Your Samples

Path B: "I want to write like [someone I admire]"

You want to emulate an author, publication, or brand voice. → If the writer is well-known (major author, publication), you may be able to use their name directly in prompts. For custom or lesser-known voices, you'll still need samples. → Proceed to Phase 1: Gather Your Samples

Path C: "I'm not sure what my voice is yet"

You want to discover and develop your voice through this process. → Start with Phase 0: Voice Discovery (below)


Phase 0: Voice Discovery (Optional)

*Skip this if you already have samples or know what voice you want.*

To discover your voice, we need to understand how you naturally communicate when you're not performing.

Answer these questions:

  1. How do you explain things to friends?

- Do you use stories and examples? - Do you build logical arguments? - Do you make jokes? - Do you ask questions to draw them in?

  1. What writers or publications do you gravitate toward?

- Not who you think you *should* read—who do you actually read? - What about their style appeals to you?

  1. What's your natural sentence length?

- Short and punchy? - Long and flowing? - Variable depending on the point?

  1. How do you feel about jargon?

- Love it (signals expertise)? - Hate it (pretentious)? - Use it sparingly when precise?

  1. What's your relationship with your reader?

- Peer/friend? - Mentor/teacher? - Curious explorer?

Based on your answers, find 2-3 pieces you've written that feel authentic. Use those for Phase 1.


Phase 1: Gather Samples

Collect 2-5 writing samples. Paste each one into a separate document, or note where they can be found.

For each sample, note:

  • Source/context (blog post, newsletter, etc.)
  • When it was written
  • Why you chose it

Phase 2: Extract Patterns

Analyze each sample across these dimensions:

2.1 Sentence Architecture

ElementWhat to Look For
LengthAverage sentence length (short/medium/long)
VariationDoes length vary deliberately for rhythm?
ComplexitySimple declarative? Compound? Complex?
PunctuationEm dashes? Semicolons? Minimal?
FragmentsUsed for emphasis? Never?

Signature structures to identify:

  • Opening patterns ("Here's the thing:")
  • Rhythm breaks ("But.")
  • List patterns ("First... Second... Third...")
  • Closing moves (questions, callbacks, calls to action)

2.2 Word Choice

ElementWhat to Look For
Vocabulary levelSimple/moderate/sophisticated
Formality1-5 scale (casual to formal)
ContractionsAlways/sometimes/never
JargonIndustry terms? Avoided?
Strong languageProfanity level/style if any

Identify:

  • Favorite words (appears 3+ times across samples)
  • Avoided words (never appears despite opportunity)
  • Signature phrases

2.3 Tone & Attitude

ElementWhat to Look For
Emotional registerWarm? Cool? Intense? Measured?
Reader relationshipPeer? Mentor? Friend? Expert?
Humor styleSarcasm? Wordplay? Self-deprecation? None?
Certainty levelDefinitive or exploratory?
Attitude toward subjectPassionate? Skeptical? Curious?

2.4 Structural Moves

ElementWhat to Look For
OpeningsStory? Question? Bold claim? Scene-setting?
TransitionsSeamless? Signposted? Abrupt?
ClosingsCall to action? Question? Summary? Callback?
Paragraph lengthShort punchy? Long flowing? Mixed?

2.5 Distinctive Techniques

What makes this voice *recognizable*?

  • Rhetorical devices (questions, analogies, callbacks)
  • Signature moves unique to this writer
  • How evidence is presented (data? anecdotes? logic?)
  • Pattern interrupts (how does the writer surprise?)

2.6 What They DON'T Do

Just as important:

  • Phrases never used
  • Structures avoided
  • Topics skipped
  • Formality levels never hit

Phase 3: Synthesize

The Voice in One Paragraph

Write a single paragraph capturing the essence:

"[Name]'s voice is [primary characteristics]. They write like [relationship to reader], using [key techniques]. Their tone is [tone], with [humor style if applicable]. Sentences tend to be [structure]. They favor [word choice] and avoid [avoidances]. The overall effect is [feeling/impact]."

Voice Spectrum

Rate the voice on these scales (1-5):

Formal ←――――――――――→ Casual         [ ]
Expert ←――――――――――→ Peer           [ ]
Serious ←―――――――――→ Playful        [ ]
Reserved ←――――――――→ Opinionated    [ ]
Abstract ←――――――――→ Concrete       [ ]

Phase 4: Generate the Voice Skill

Using your analysis, create voice-[name].skill.md with this structure:

---
name: voice-[name]
description: Write in [Name]'s distinctive voice. [One sentence characterization]. Use with anti-ai-writing for authentic content.
---

# Voice: [Name]

[Your one-paragraph voice description from Phase 3]

## Voice Spectrum

- Formal/Casual: [1-5]
- Expert/Peer: [1-5]
- Serious/Playful: [1-5]
- Reserved/Opinionated: [1-5]
- Abstract/Concrete: [1-5]

## Core Characteristics

### Sentence Architecture
[Your findings from 2.1]

### Word Choice
[Your findings from 2.2]

### Tone & Attitude
[Your findings from 2.3]

### Structural Moves
[Your findings from 2.4]

### Distinctive Techniques
[Your findings from 2.5]

### Avoidances
[Your findings from 2.6]

## Example Patterns

### Signature Openings
[3-5 examples from samples with pattern explanation]

### Signature Transitions
[3-5 examples]

### Signature Closings
[3-5 examples]

### Signature Sentences
[5-10 exemplary sentences that capture the voice]

## The [Name] Test

Before publishing, ask:
- [ ] Would [Name] actually write this sentence?
- [ ] Does it match the voice spectrum above?
- [ ] Are signature patterns present?
- [ ] Are avoidances absent?
- [ ] Does it *feel* right?

## Anti-Patterns

If you see these, you've drifted from the voice:
- [List 5-10 patterns that would violate this voice]

Phase 5: Validate

Test your voice skill:

  1. Write a short piece using only the voice skill
  2. Compare to original samples
  3. Identify gaps or mischaracterizations
  4. Refine based on testing

Validation checklist:

  • Voice description captures the essence
  • Spectrum ratings feel accurate
  • Example patterns are genuinely representative
  • Test output sounds like the samples
  • Nothing important was missed

Human Writing Fundamentals

*Every voice skill should build on these principles from anti-ai-writing.*

The Energy Transfer Principle

The best writing is a transfer of energy from writer to reader. When analyzing samples, notice how the writer transfers energy:

  • Do they write conversations or speeches?
  • Do they speak WITH their audience or AT them?
  • Where do they use specific, concrete language vs. abstract ideas?

The SUCKS Framework

Apply this when generating voice output:

S - Specific: Who is the ONE reader? U - Unique & Useful: Does it change how they think, feel, or act? C - Clear, Curious, Conversational: Does it read like talking to a friend? K - Kept Simple & Structured: Simple ideas, clear structure? S - Sticky: Are there memorable phrases they'll repeat?

Sticky Sentence Techniques

When identifying signature sentences in Phase 2, look for these techniques:

Alliteration — Same starting sounds

  • "Specificity is the secret"
  • "The best jobs are neither decreed nor degreed"

Symmetry — Parallel structure

  • "Read for awareness. Write for understanding."
  • "It's not 10,000 hours. It's 10,000 iterations."

Contrast — Opposing ideas

  • "To be everywhere is to be nowhere."
  • "Be clear, not clever. Concise, not complex."

Rhythm — Pleasing cadence

  • Sentence length variation for effect
  • Short sentences for emphasis
  • Longer sentences for flow

Include the writer's use of these techniques in your voice skill.

AI Tells to Eliminate

When using your voice skill, watch for these patterns that signal AI involvement:

The Correlative Construction (most common):

  • ❌ "X aren't just Y - they're Z"
  • ❌ "It's not about X, it's about Y"

Forbidden Openers:

  • ❌ "In the ever-evolving world of..."
  • ❌ "Gone are the days when..."
  • ❌ "Let that sink in"

Hedging Language:

  • ❌ "This might help you" → "This will help you"
  • ❌ "It could be argued..." → state it directly

Overused Softeners:

  • Too much "just" and "actually"
  • Passive voice ("was determined")
  • Corporate jargon

Test: For each sentence, ask: Would this appear in ChatGPT output? If yes, rewrite it.


Using Your Voice Skill

Save to: .claude/skills/voice-[name]/SKILL.md

Invoke alongside other skills:

  • voice-[name] + anti-ai-writing = Authentic, humanized content
  • voice-[name] + ghostwriter = Long-form pieces in voice
  • voice-[name] + social-content-creation = Platform-specific posts

Update as needed: Your voice evolves. Revisit the skill quarterly or when something feels off.


Related Skills

  • anti-ai-writing — Core humanization engine (use with all voice work)
  • ghostwriter — For long-form content in someone else's voice
  • transcript-polisher — For interview-based content

*One well-crafted voice skill compounds across everything you create.*

平台分布

Codex

35.6%
按下载量换算45

Claude

31.69%
按下载量换算40

Cursor

19.64%
按下载量换算25

Gemini CLI

10.38%
按下载量换算13

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

权限未标注

当前来源未能明确判断权限范围,默认进入异常复核队列。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills