Back to Details

voice-validator

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Voice Validator Skill

Voice Validator Skill

Operator Context

操作者背景

This skill operates as an operator for voice validation workflows, configuring Claude's behavior for rigorous critique-and-rewrite enforcement. It implements the Iterative Refinement architectural pattern -- scan content, identify violations, revise, rescan -- with Domain Intelligence embedded in voice-specific negative prompt checklists and pass/fail criteria.
该Skill作为语音验证工作流的操作者，配置Claude的行为以严格执行批评-重写循环。它采用Iterative Refinement（迭代优化）架构模式——扫描内容、识别违规、修改、重新扫描——并将领域智能嵌入到特定语音风格的负面提示清单和通过/失败判定标准中。

Hardcoded Behaviors (Always Apply)

硬编码行为（始终生效）

  • CLAUDE.md Compliance: Read and follow repository CLAUDE.md before validating
  • Over-Engineering Prevention: Fix only voice violations. No content rewriting beyond voice fidelity
  • Scan Before Revise: NEVER revise content without first scanning against the full checklist
  • Evidence Required: Every violation must cite a specific quote from the content
  • Maximum 3 Iterations: If still failing after 3 rewrites, output with flagged concerns
  • Preserve Intent: Revisions fix voice violations only -- never alter meaning or substance
  • CLAUDE.md合规性：验证前请阅读并遵循仓库中的CLAUDE.md
  • 防止过度设计：仅修复语音风格违规问题，除语音保真度外，不得修改内容其他部分
  • 先扫描后修改：在未针对完整清单扫描前，绝对不要修改内容
  • 需提供证据：每一处违规都必须引用内容中的特定原文
  • 最多3次迭代：若3次重写后仍未通过，输出带有标记问题的内容
  • 保留意图：修改仅修复语音风格违规——绝不能改变内容的含义或实质

Default Behaviors (ON unless disabled)

默认行为（默认开启，可关闭）

  • Full Checklist Scan: Run all categories (tone, structure, sentence, language, emotion, questions, metaphors)
  • Violation Report: Output structured validation report with quoted violations and fixes
  • Auto-Revise on Fail: Automatically produce revised version when violations detected
  • Rescan After Revision: Re-run checklist against revised content to confirm pass
  • LinkedIn Test: Apply quick "could this be posted on LinkedIn without edits?" heuristic
  • Mode Detection: Identify voice mode (1-5) for mode-specific checks
  • 完整清单扫描：运行所有类别（语气、结构、句式、语言、情感、提问方式、隐喻）的扫描
  • 违规报告：输出结构化验证报告，包含引用的违规内容及修复方案
  • 失败时自动修改：检测到违规时自动生成修改后的版本
  • 修改后重新扫描：针对修改后的内容重新运行清单以确认通过
  • LinkedIn测试：应用快速启发式检查——「无需修改即可发布到LinkedIn？」
  • 模式检测：识别语音模式（1-5）以进行模式专属检查

Optional Behaviors (OFF unless enabled)

可选行为（默认关闭，可开启）

  • Script Validation: Run 
    voice_validator.py
    for deterministic banned-pattern detection
  • Inline Validation: Silent self-check within conversation without full report
  • Cross-Voice Comparison: Compare output against wrong-voice patterns to verify distinctness
  • 脚本验证：运行
    voice_validator.py
    进行确定性禁用模式检测
  • 内联验证：在对话中静默自检，不生成完整报告
  • 跨语音对比：将输出与错误语音风格的模式对比，验证差异性

What This Skill CAN Do

该Skill可实现的功能

  • Validate content against voice-specific negative prompt checklists
  • Identify specific violations with quoted evidence and category labels
  • Revise content to fix voice violations while preserving intent
  • Enforce iterative scan-revise-rescan loops up to 3 iterations
  • Distinguish between different voice profiles with mode-specific criteria
  • 根据特定语音风格的负面提示清单验证内容
  • 引用证据并标记类别，识别具体违规问题
  • 在保留内容意图的前提下，修改内容以修复语音风格违规
  • 强制执行扫描-修改-重新扫描的迭代循环，最多3次
  • 通过模式专属标准区分不同语音配置文件

What This Skill CANNOT Do

该Skill不可实现的功能

  • Generate content in a target voice from scratch (use the appropriate voice skill instead)
  • Create or modify voice profiles (use voice_analyzer.py instead)
  • Edit content for non-voice concerns like grammar or accuracy (use anti-ai-editor instead)
  • Skip the scan phase and go straight to rewriting
  • Validate voices that have no defined checklist
  • 从零开始生成目标语音风格的内容（请使用对应的语音Skill）
  • 创建或修改语音配置文件（请使用voice_analyzer.py）
  • 针对非语音相关问题编辑内容，如语法或准确性（请使用anti-ai-editor）
  • 跳过扫描阶段直接进行重写
  • 验证无定义清单的语音风格

Instructions

操作步骤

Phase 1: IDENTIFY TARGET

阶段1：确定目标

Goal: Determine the voice, mode, and content to validate.
Step 1: Identify voice target
  • Determine target voice from context or user instruction
  • Identify mode if applicable -- casual modes may have additional specific checks
Step 2: Load content
  • Read the content to validate
  • Note content length -- longer content is more prone to drift
Gate: Voice target and mode identified. Content loaded. Proceed only when gate passes.
目标：确定要验证的语音风格、模式和内容。
步骤1：确定目标语音风格
  • 从上下文或用户指令中确定目标语音风格
  • 若适用，确定模式——非正式模式可能有额外的专属检查
步骤2：加载内容
  • 读取待验证的内容
  • 记录内容长度——长内容更易出现语音漂移
准入条件：已确定目标语音风格和模式，已加载内容。仅当满足条件时方可继续。

Phase 2: SCAN

阶段2：扫描

Goal: Run full checklist against content and identify all violations.
Step 1: Run negative prompt checklist
Check all categories against the target voice's checklist:
  • Tone: Does the tone match the voice profile? (e.g., too polished, too corporate, missing warmth)
  • Structure: Does the structure match? (e.g., front-loaded constraints, clean outlines, wrap-ups)
  • Sentences: Do sentence patterns match? (e.g., dramatic short sentences, rhetorical flourishes, symmetrical structure)
  • Language: Any banned words? (amazing, terrible, revolutionary, perfect, game-changing, transformative, incredible, outstanding, exceptional, groundbreaking), marketing/hype, inspirational, unnecessary superlatives
  • Emotion: Does emotion handling match? (e.g., explicitly named emotions, venting/ranting, moralizing)
  • Questions: Do question patterns match? (e.g., open-ended brainstorming, vague curiosity)
  • Metaphors: Do metaphor patterns match? (e.g., journey/path, biological/growth, narrative/story)
Step 2: Check pass conditions
Verify the content matches the target voice's positive identity markers. Common pass conditions include:
  • Feels like the person actually wrote it
  • Voice-specific patterns are present (thinking out loud, warmth, precision, etc.)
  • Could NOT be posted on LinkedIn without edits (for casual voices)
  • Does NOT sound like AI wrote it
  • Mode-specific patterns are present (casual modes: no preamble, no wrap-up; formal modes: structured flow)
Step 3: Document violations
For each violation, record:
  1. Category (tone, structure, sentence, language, emotion, question, metaphor)
  2. Quoted text from the content
  3. Specific fix recommendation
Gate: Full checklist scanned. All violations documented with evidence. Proceed only when gate passes.
目标：针对内容运行完整清单，识别所有违规问题。
步骤1：运行负面提示清单
针对目标语音风格的清单，检查所有类别：
  • 语气：语气是否匹配语音配置文件？（例如：过于正式、过于商业化、缺乏温度）
  • 结构：结构是否匹配？（例如：前置约束、清晰大纲、总结收尾）
  • 句式：句式模式是否匹配？（例如：戏剧性短句、修辞修饰、对称结构）
  • 语言：是否存在禁用词汇？（amazing、terrible、revolutionary、perfect、game-changing、transformative、incredible、outstanding、exceptional、groundbreaking）、营销/炒作话术、励志类表述、不必要的夸张词汇
  • 情感：情感处理方式是否匹配？（例如：明确命名情感、宣泄/吐槽、说教）
  • 提问方式：提问模式是否匹配？（例如：开放式头脑风暴、模糊的好奇心）
  • 隐喻：隐喻模式是否匹配？（例如：旅程/路径、生物/成长、叙事/故事类隐喻）
步骤2：检查通过条件
验证内容是否匹配目标语音风格的正向标识。常见通过条件包括：
  • 感觉像是目标人物实际撰写的内容
  • 存在语音专属模式（例如：出声思考、温暖感、精准性等）
  • （针对非正式语音）无需修改即可发布到LinkedIn则不通过
  • 听起来不像AI生成的内容
  • 存在模式专属特征（非正式模式：无开场白、无收尾；正式模式：结构化流程）
步骤3：记录违规问题
针对每一处违规，记录：
  1. 类别（语气、结构、句式、语言、情感、提问方式、隐喻）
  2. 内容中引用的违规原文
  3. 具体修复建议
准入条件：已完成完整清单扫描，所有违规问题均已记录并附证据。仅当满足条件时方可继续。

Phase 3: REVISE

阶段3：修改

Goal: Fix all violations while preserving content intent and substance.
Step 1: Apply fixes
  • Address each violation with the smallest change that resolves it
  • Preserve the original meaning and information
  • Maintain natural flow -- fixes should not create new violations
Step 2: Verify no overcorrection
  • Ensure revisions did not strip necessary content
  • Confirm the substance and technical accuracy remain intact
Gate: All documented violations addressed. Intent preserved. Proceed only when gate passes.
目标：修复所有违规问题，同时保留内容的意图和实质。
步骤1：应用修复方案
  • 采用最小改动解决每一处违规问题
  • 保留原有的含义和信息
  • 保持自然流畅——修复不得产生新的违规
步骤2：验证无过度修改
  • 确保修改未删除必要内容
  • 确认内容的实质和技术准确性未受影响
准入条件：所有已记录的违规问题均已处理，内容意图得以保留。仅当满足条件时方可继续。

Phase 4: VERIFY

阶段4：验证

Goal: Confirm revised content passes all checks.
Step 1: Rescan revised content
Run the full checklist from Phase 2 against the revised version.
Step 2: Evaluate result
  • If PASS: Output final content with validation report
  • If FAIL and iteration < 3: Return to Phase 3 with new violations
  • If FAIL and iteration = 3: Output content with flagged remaining concerns
Step 3: Output validation report
VOICE VALIDATION: [Voice Name] Mode [mode]
SCAN RESULT: [PASS/FAIL]
VIOLATIONS DETECTED: [N]
ITERATION: [1-3]

[If violations:]
1. [Category]: "[quoted violation]"
   Fix: [specific correction]

2. [Category]: "[quoted violation]"
   Fix: [specific correction]

REVISED OUTPUT:
[Corrected content]

RESCAN RESULT: [PASS/FAIL]
Gate: Content passes all checks, or maximum iterations reached with flagged concerns. Validation complete.
目标：确认修改后的内容通过所有检查。
步骤1：重新扫描修改后的内容
针对修改后的版本运行阶段2的完整清单。
步骤2：评估结果
  • 若通过：输出最终内容及验证报告
  • 若未通过且迭代次数<3：带着新的违规问题返回阶段3
  • 若未通过且迭代次数=3：输出带有标记剩余问题的内容
步骤3：输出验证报告
VOICE VALIDATION: [语音名称] 模式 [mode]
SCAN RESULT: [通过/未通过]
VIOLATIONS DETECTED: [数量]
ITERATION: [1-3]

[若存在违规：]
1. [类别]: "[引用的违规内容]"
   修复方案: [具体修正内容]

2. [类别]: "[引用的违规内容]"
   修复方案: [具体修正内容]

修改后的输出:
[修正后的内容]

重新扫描结果: [通过/未通过]
准入条件：内容通过所有检查，或已达到最大迭代次数并标记剩余问题。验证完成。

Examples

示例

Example 1: Technical Voice Validation

示例1：技术语音风格验证

User says: "Validate this draft is in the right voice" Actions:
  1. Identify target voice from context, determine mode from content style (IDENTIFY TARGET)
  2. Run full 7-category negative prompt checklist, find 2 violations (SCAN)
  3. Fix "I'm excited to share" (named emotion) and "This changes everything" (dramatic short sentence) (REVISE)
  4. Rescan revised content, confirm PASS (VERIFY) Result: Clean content with validation report
用户指令：「验证这份草稿是否符合正确的语音风格」 操作：
  1. 从上下文中确定目标语音风格，从内容风格中确定模式（确定目标）
  2. 运行7类完整负面提示清单，发现2处违规（扫描）
  3. 修复「I'm excited to share」（明确命名情感）和「This changes everything」（戏剧性短句）（修改）
  4. 重新扫描修改后的内容，确认通过（验证） 结果：符合要求的内容及验证报告

Example 2: Community Voice Validation

示例2：社区语音风格验证

User says: "Does this sound like the right voice?" Actions:
  1. Identify target voice from context (IDENTIFY TARGET)
  2. Scan against voice checklist, find missing warmth and no sensory details (SCAN)
  3. Add experiential language and warmth while preserving substance (REVISE)
  4. Rescan, confirm warmth and sensory details present, PASS (VERIFY) Result: Content matches voice profile
用户指令：「这听起来是正确的语音风格吗？」 操作：
  1. 从上下文中确定目标语音风格（确定目标）
  2. 针对语音清单扫描，发现缺乏温暖感且无感官细节（扫描）
  3. 在保留内容实质的前提下，添加体验式语言和温暖感（修改）
  4. 重新扫描，确认温暖感和感官细节已存在，通过（验证） 结果：内容匹配语音配置文件

Error Handling

错误处理

Error: "Voice Target Unclear"

错误：「目标语音风格不明确」

Cause: Content doesn't specify which voice to validate against, or context is ambiguous Solution:
  1. Check conversation context for voice mentions
  2. Look for voice-specific patterns to infer target
  3. If still unclear, ask user to specify voice name and mode
原因：内容未指定要验证的语音风格，或上下文模糊 解决方案：
  1. 检查对话上下文是否提及语音风格
  2. 寻找语音专属模式以推断目标
  3. 若仍不明确，请求用户指定语音名称和模式

Error: "Violations Persist After 3 Iterations"

错误：「3次迭代后违规仍存在」

Cause: Fundamental mismatch between content substance and voice requirements, or conflicting checklist items Solution:
  1. Output content with clearly flagged remaining violations
  2. List specific checklist items that resist correction
  3. Suggest the content may need to be regenerated from scratch with the correct voice skill
原因：内容实质与语音风格要求存在根本性不匹配，或清单项存在冲突 解决方案：
  1. 输出带有明确标记剩余违规的内容
  2. 列出难以修正的具体清单项
  3. 建议可能需要使用正确的语音Skill从头重新生成内容

Error: "Revision Introduced New Violations"

错误：「修改引入新的违规」

Cause: Fixing one category created violations in another (e.g., removing dramatic sentences introduced polished phrasing) Solution:
  1. Address new violations in next iteration
  2. If oscillating between two violation types, fix both simultaneously
  3. Prioritize tone and language violations over structural ones
原因：修复某一类别的违规导致另一类别出现违规（例如：删除戏剧性短句后引入了正式措辞） 解决方案：
  1. 在下次迭代中处理新的违规
  2. 若在两类违规间反复切换，同时修复两类问题
  3. 优先处理语气和语言类违规，而非结构类

Anti-Patterns

反模式

Anti-Pattern 1: Revising Without Scanning

反模式1：未扫描直接修改

What it looks like: "This doesn't sound right, let me rewrite it" without running the checklist Why wrong: Subjective assessment misses specific violations. May "fix" things that aren't broken while missing real issues. Do instead: Complete Phase 2 scan with documented violations before any revision.
表现：「这听起来不对，我来重写」，未运行清单扫描 问题：主观评估会遗漏具体违规，可能「修复」原本没问题的内容，同时忽略真正的问题 正确做法：完成阶段2的扫描并记录违规后，再进行任何修改

Anti-Pattern 2: Over-Revising Beyond Voice

反模式2：修改超出语音风格范围

What it looks like: Rewriting entire paragraphs, changing arguments, adding new points during voice correction Why wrong: Voice validation fixes voice only. Changing substance is scope creep that alters the author's intent. Do instead: Make the smallest change that resolves each voice violation. Preserve all meaning.
表现：在语音风格修正过程中重写整个段落、改变论点、添加新内容 问题：语音验证仅修复语音风格问题，修改内容实质属于范围蔓延，会改变作者的意图 正确做法：采用最小改动解决每一处语音风格违规，保留所有原有含义

Anti-Pattern 3: Skipping the Rescan

反模式3：跳过重新扫描

What it looks like: "I fixed the violations, it should be fine now" without re-running the checklist Why wrong: Fixes can introduce new violations. "Should be fine" is a rationalization. Do instead: Always run Phase 4 rescan. Every revision gets a full checklist pass.
表现：「我已经修复了违规，应该没问题了」，未重新运行清单 问题：修复可能引入新的违规，「应该没问题」是主观合理化 正确做法：始终运行阶段4的重新扫描，每一次修改都要通过完整清单检查

Anti-Pattern 4: Passing Content That Sounds Like LinkedIn

反模式4：通过了像LinkedIn风格的内容

What it looks like: Content is polished, quotable, and shareable -- but marked as PASS Why wrong: The LinkedIn test catches ~80% of voice violations. If it reads well on LinkedIn, it fails the target voice. Do instead: Apply the quick check: "Could this be posted on LinkedIn without edits?" If yes, it FAILS.
表现：内容正式、适合引用和分享——却被标记为通过 问题：LinkedIn测试可检测约80%的语音风格违规，若内容适合发布到LinkedIn，则不符合目标语音风格 正确做法：应用快速检查——「无需修改即可发布到LinkedIn？」若答案是，则判定为未通过

References

参考资料

This skill uses these shared patterns:
  • Anti-Rationalization - Prevents shortcut rationalizations
  • Verification Checklist - Pre-completion checks
该Skill使用以下共享模式：
  • Anti-Rationalization - 防止主观合理化捷径
  • Verification Checklist - 完成前检查

Domain-Specific Anti-Rationalization

领域专属反合理化

RationalizationWhy It's WrongRequired Action
"It sounds close enough"Close enough ≠ voice fidelityRun full checklist, fix all violations
"Only one small violation"One violation breaks immersionFix it. No exceptions
"The substance matters more than voice"Voice IS the deliverable in this contextComplete all 4 phases
"I already know what's wrong"Knowing ≠ documenting with evidenceScan and cite specific quotes
合理化借口问题所在要求操作
「听起来差不多了」差不多≠语音保真运行完整清单，修复所有违规
「只有一处小违规」一处违规即可打破沉浸感修复该问题，无例外
「内容实质比语音风格更重要」在此场景下，语音风格就是交付成果完成全部4个阶段
「我已经知道问题在哪了」知道≠引用证据记录扫描并引用具体原文

Related Skills

相关Skill

  • voice-{name}
    - Generates content in a specific voice (validate output with this skill)
  • anti-ai-editor
    - Complementary anti-AI pattern detection
  • voice-orchestrator
    - Multi-step voice generation pipeline that invokes this skill
  • voice-{name}
    - 生成特定语音风格的内容（可使用本Skill验证输出）
  • anti-ai-editor
    - 互补的AI模式检测工具
  • voice-orchestrator
    - 多步骤语音生成流水线，会调用本Skill",