quality-review

Quality Review - Quality Review

Trigger Commands: /review Trigger Keywords: review, proofread, check, audit, version evaluation, A/B testing, version comparison

Introduction

Quality review skill suite, integrating two major functions: text review and version evaluation. Text review provides comprehensive grammar, logic, and compliance checks; version evaluation supports multi-version comparative analysis and A/B testing decision support.

Function 1: Review Text

Review Text - Text Review

Introduction

Conduct a comprehensive review of the text, including grammar, spelling, punctuation, logic, clarity, and other aspects, and provide improvement suggestions. After the review is completed, an AI disclaimer will be added at the end of the output. After generating suggestions, humanize_text is automatically called to remove AI traces.

Use Cases

Perform final review of the draft
Find and correct text errors
Improve text quality
Ensure the accuracy and professionalism of the content
Final check before publication

Input Parameters

Required Parameters

source_text (string): The original text to be reviewed

Optional Parameters

review_aspects (string[]): Review focus (multiple selections allowed)
- grammar - Grammar check
- spelling - Spelling check
- punctuation - Punctuation check
- logic - Logic check
- clarity - Clarity of expression check
- consistency - Consistency check
- factual_accuracy - Factual accuracy check
- compliance - Compliance check (check sensitive words and violating content)
- Default value: all checks
target_audience (string): Target audience
- general - General public
- professional - Professionals
- academic - Academia
- youth - Young people
- business - Business professionals
- Default value: general
tone_requirement (string): Tone requirement
- formal - Formal
- semi-formal - Semi-formal
- casual - Informal
- Default value: semi-formal
return_format (string): Return format
- summary_only - Return only summary and revision suggestions
- detailed - Return detailed analysis and revision suggestions
- with_marked_text - Return annotated original text and revised version
- Default value: detailed
highlight_changes (boolean): Whether to highlight modifications
- true - Mark modified content with special symbols
- false - Do not highlight
- Default value: true

Output

Review results, including problem analysis, revision suggestions, revised version, and the AI disclaimer "The above content has been reviewed and optimized by a large AI model and must be manually reviewed again before use." Revision suggestions will automatically call humanize_text to remove AI traces.

Processing Rules

Review Dimensions

1. Grammar check (grammar)

Subject-verb agreement check
Tense consistency check
Word order reasonableness check
Grammatical structure completeness check

2. Spelling check (spelling)

Typo check
Homophone misuse check
Easily confused character check
Technical term spelling check

3. Punctuation check (punctuation)

Punctuation usage standard check
Mixed use of Chinese and English punctuation check
Punctuation position check
Punctuation completeness check

4. Logic check (logic)

Argument logic check
Causal relationship check
Consistency check
Reasoning soundness check

5. Clarity check (clarity)

Ambiguous expression check
Redundant expression check
Information overload check
Expression accuracy check

6. Consistency check (consistency)

Terminology consistency check
Person consistency check
Tense consistency check
Format consistency check

7. Factual accuracy check (factual_accuracy)

Data accuracy check
Date and time check
Proper noun check
Citation accuracy check

8. Compliance check (compliance)

Sensitive word check
Violating content check
Copyrighted content check
False information check

Issue severity classification

1. Critical issues (Critical)

Errors that severely affect understanding
Obvious factual errors
Serious compliance issues
Must be revised

2. Important issues (Important)

Errors that affect reading fluency
Expressions that are not accurate enough
Terminology that is not consistent enough
Revision recommended

3. Minor issues (Minor)

Minor expression flaws
Expressions with room for optimization
Wording that is not perfect enough
Optional revision

Review process

1. Preliminary scan

Quickly scan the text to identify obvious issues
Count the number of issues and their type distribution
Determine review priorities

2. In-depth analysis

Analyze paragraph by paragraph in depth
Discover potential issues
Record issue locations and types

3. Issue classification

Classify issues by severity
Mark issue locations
Provide issue descriptions

4. Revision suggestions

Provide a revision plan for each issue
Explain the reasons for revision
Give revision suggestions

5. Overall assessment

Assess the overall quality of the text
Give directions for improvement
Provide summary suggestions

Revision suggestion format

Summary format (summary_only)

【审校摘要】
原文质量评分：X/10
发现问题总数：X个（严重：X个，重要：X个，一般：X个）

【主要问题】
1. XXX问题
2. XXX问题
3. XXX问题

【修订建议】
（概括性建议）

【修订版本】
（整体修订后的文本）

Detailed format (detailed)

【审校摘要】
原文质量评分：X/10
发现问题总数：X个（严重：X个，重要：X个，一般：X个）

【详细问题分析】
1. [严重] 语法错误
   位置：第X行第X字
   问题：XXX
   建议：XXX
   修订：XXX

2. [重要] 表达不够清晰
   位置：第X行第X字
   问题：XXX
   建议：XXX
   修订：XXX

...

【修订版本】
（整体修订后的文本，用特殊符号标注修改）

【改进方向】
（进一步优化的建议）

Marked format (with_marked_text)

【审校摘要】
...

【标注原文】
（用[[删除内容]]表示删除，用【新增内容】表示新增）

【修订版本】
（清晰的修订后文本）

【问题清单】
1. XXX问题 → 已修正
2. XXX问题 → 已修正
...

Highlighting rules

Deleted content

Use[[内容]]to mark content to be deleted
For example:这个[[非常]]好的产品

Added content

Use【内容】to mark content to be added
For example:这个产品【质量】很好

Modified content

Use[[原内容]]→【新内容】to mark modifications
For example:[[非常]]→【特别】好的产品

Quick mode

When the user input contains a complete source_text, execute directly using the default parameters (all checks, general public audience, semi-formal tone, detailed format, highlighting).

Interactive question template (only displayed when the user does not provide the necessary parameters)

我将为您审校这段文本，请确认以下要求：

1. 审校重点（可多选，可选）：
   [1] 语法检查
   [2] 拼写检查
   [3] 标点符号检查
   [4] 逻辑检查
   [5] 表达清晰度检查
   [6] 一致性检查
   [7] 事实准确性检查
   [8] 合规性检查（敏感词、违规内容）

2. 目标受众（可选）：
   [1] 普通大众
   [2] 专业人士
   [3] 学术界
   [4] 年轻群体
   [5] 商务人士

3. 语气要求（可选）：
   [1] 正式
   [2] 半正式（推荐）
   [3] 非正式

4. 返回格式（可选）：
   [1] 仅返回摘要和修订建议
   [2] 返回详细分析和修订建议（推荐）
   [3] 返回标注原文和修订版本

5. 是否高亮显示修改（可选）：
   [1] 高亮显示修改内容（推荐）
   [2] 不高亮

您可以：
- 一次性回复所有选择（如：1,2,3,4,5,6,7,8，1，2，2，1）
- 回复部分参数，我再确认是否需要补充
- 回复"审校"或"好的"，使用推荐设置执行（推荐：全部检查，普通大众受众，半正式语气，详细格式，高亮显示）
- 如果你有其他想法，可在回复的时候一并描述告知

请回复：

Example

Example 1: Comprehensive proofreading

Input：

source_text: "这个产品有很多好的功能，能够帮助用户提高工作的效率。但是有时候它会有些问题，需要进一步改进。我们相信，通过不断的努力，这个产品会越来越好。"
review_aspects: ["grammar", "spelling", "clarity", "consistency"]
return_format: "detailed"

Output：

【审校摘要】
原文质量评分：7/10
发现问题总数：3个（严重：0个，重要：2个，一般：1个）

【详细问题分析】
1. [重要] 表达不够简洁
   位置：第1句
   问题："有很多好的功能"表达略显啰嗦
   建议：改为"功能丰富"或"功能多样"
   修订：这个产品功能丰富，能够帮助用户提高工作效率。

2. [重要] 表达不够明确
   位置：第2句
   问题："有时候它会有些问题"表述模糊，未说明具体问题
   建议：明确指出具体问题类型
   修订：但在特定情况下可能出现兼容性问题，需要进一步改进。

3. [一般] 逻辑可优化
   位置：第3句
   问题："通过不断的努力"稍显空泛
   建议：可以更具体地说明努力方向
   修订：我们相信，通过持续优化和用户反馈，这个产品会越来越好。

【修订版本】
这个产品功能丰富，能够帮助用户提高工作效率。但在特定情况下可能出现兼容性问题，需要进一步改进。我们相信，通过持续优化和用户反馈，这个产品会越来越好。

【改进方向】
1. 可以进一步丰富内容，加入具体案例
2. 可以增加数据支撑，提升说服力
3. 可以优化段落结构，使内容更清晰

---

以上内容为AI大模型审校优化，在使用前需再人工审核

Example 2: Compliance check

Input：

source_text: "这个投资产品绝对没有任何风险，保证100%的收益，没有任何损失。"
review_aspects: ["compliance", "factual_accuracy"]
return_format: "summary_only"

Output：

【审校摘要】
原文质量评分：3/10
发现问题总数：3个（严重：3个，重要：0个，一般：0个）

【主要问题】
1. [严重] 合规问题：使用"绝对没有任何风险"等绝对化表述，违反金融广告规范
2. [严重] 合规问题：使用"保证100%的收益"等承诺性表述，违反金融广告规范
3. [严重] 事实问题：没有任何投资产品能够保证100%的收益和零风险

【修订建议】
强烈建议修改所有绝对化表述和承诺性表述，改为更加客观、谨慎的表达方式。金融产品宣传应遵循"风险提示"原则，充分揭示风险。

【修订版本】
这个投资产品具有一定的收益潜力，但也存在相应的投资风险。投资前请仔细阅读产品说明书，充分了解产品特点和风险等级，根据自身风险承受能力进行投资。

---

以上内容为AI大模型审校优化，在使用前需再人工审核

Example 3: Detailed format + highlighting

Input：

source_text: "我们做了一个研究，想看看这个方法到底好不好用。结果显示，这个方法比其他方法都好。"
review_aspects: ["grammar", "clarity", "tone_requirement"]
tone_requirement: "formal"
return_format: "detailed"
highlight_changes: true

Output：

【审校摘要】
原文质量评分：6/10
发现问题总数：3个（严重：0个，重要：3个，一般：0个）

【详细问题分析】
1. [重要] 表达不够正式
   位置：第1句
   问题："做了一个研究"、"想看看"、"到底好不好用"表达过于口语化
   建议：使用更正式的学术用语
   修订：【本研究旨在】评估【该方法的】有效性

2. [重要] 表达不够准确
   位置：第2句
   问题："比其他方法都好"表述过于绝对
   建议：使用更准确的表达
   修订：研究结果表明，该方法【在多个指标上】优于【现有】其他方法

3. [重要] 结构需要优化
   位置：整体
   问题：两句话衔接不够紧密
   建议：增加连接词，使逻辑更清晰
   修订：【本研究旨在评估该方法的】有效性。研究结果表明，该方法在多个指标上优于现有其他方法。

【修订版本】
【本研究旨在评估该方法的】有效性。研究结果表明，该方法在多个指标上优于现有其他方法。

【改进方向】
1. 可以进一步增加研究背景和意义
2. 可以提供具体的数据和指标
3. 可以增加对结果的详细讨论

---

以上内容为AI大模型审校优化，在使用前需再人工审核

Notes

Proofreading will automatically add an AI disclaimer at the end of the output: "The above content is AI large model proofreading optimization and must be manually reviewed before use"
Proofreading revision suggestions will automatically call humanize_text, so there is no need to manually remove the AI flavor again
Compliance checks will refer to the rules in config/compliance_rules.json and config/risk_keywords.json
For factual accuracy checks, if it cannot be confirmed, it will be marked "requires manual verification"
Highlighting mode helps users quickly locate modified content
Proofreading suggestions are for reference only, and users should decide whether to adopt them based on the actual situation

Parameter adjustment tips

After generating content, if you are not satisfied with the current result, you can try adjusting the following parameters:

💡 Parameter adjustment options:

Proofreading focus: grammarGrammar check | spellingSpelling check | punctuationPunctuation check | logicLogic check | clarityClarity check | consistencyConsistency check | factual_accuracyFactual accuracy check | complianceCompliance check (sensitive words, violating content)
Target audience: generalGeneral public | professionalProfessionals | academicAcademia | youthYoung people | businessBusiness professionals
Tone requirement: formalFormal | semi-formalSemi-formal | casualInformal
Return format: summary_onlyReturn only summary and revision suggestions | detailedReturn detailed analysis and revision suggestions | with_marked_textReturn annotated original text and revised version
Whether to highlight modifications: trueHighlight modified content | falseDo not highlight

How to use: Reply directly, such as "focus on checking logic and clarity" or "return detailed analysis"

Relationship with other skills

humanize_text: After revision suggestions are generated, this skill is automatically called to remove AI traces
polish_text: If the goal is to improve expression quality rather than find errors, polish_text should be used
check_style_consistency: If an in-depth check of style consistency is needed, this skill should be used
rewrite_content: If existing text needs to be rewritten, it can be used together

Configuration file references

intent_keywords: ["review", "proofread", "check", "audit", "correct", "error correction"]
compliance_rules: Define compliance check rules in config/compliance_rules.json
risk_keywords: Define sensitive words and risk word libraries in config/risk_keywords.json

Feature 2: Evaluate Versions

Evaluate Versions - Version evaluation

Introduction

Compare and evaluate multiple versions of the same content, analyze the advantages and disadvantages of each version, and provide version selection suggestions and optimization directions. Suitable for scenarios such as A/B testing, version comparison, and decision support.

Use cases

A/B test result analysis
Multi-version copy comparison and evaluation
Selection decisions among different options
Guidance for content optimization direction
Version iteration effect evaluation

Input parameters

Required parameters

versions (array): version list

Each version includes: version_name (version name), content (content), version_info (version information, optional)
Supports 2-5 versions

Example:

[
  {"version_name": "版本A", "content": "内容A", "version_info": "简洁版"},
  {"version_name": "版本B", "content": "内容B", "version_info": "详细版"}
]

Optional parameters

evaluation_criteria (string[]): evaluation criteria (multiple selections allowed)
- clarity - clarity
- engagement - engagement
- persuasion - persuasiveness
- professionalism - professionalism
- creativity - creativity
- readability - readability
- emotional_impact - emotional impact
- target_audience_fit - target audience fit
- Default value: all evaluations
target_audience (string): target audience
- general - general public
- professional - professionals
- youth - young people
- business - business professionals
- academic - academia
- custom description
- Default value: general
use_case (string): use case
- marketing - marketing promotion
- information - information delivery
- education - education and training
- entertainment - entertainment
- brand_building - brand building
- custom - custom
- Default value: automatically determined based on content
evaluation_method (string): evaluation method
- comparative - comparative evaluation
- scoring - scoring evaluation
- both - comparison + scoring (recommended)
- Default value: both
provide_optimization_suggestions (boolean): whether to provide optimization suggestions
- true - provide optimization suggestions for each version (recommended)
- false - do not provide
- Default value: true

Output

Version evaluation report, including detailed analysis of each version, comparative evaluation, score ranking, selection suggestions, optimization directions, etc.

Processing rules

Detailed explanation of evaluation criteria

1. clarity

Evaluation points:
- Whether the information is expressed clearly and explicitly
- Whether the logic is smooth
- Whether there are ambiguous expressions
- Whether the information density is reasonable
Scoring criteria (0-10 points):
- 9-10 points: expression is very clear, no ambiguity
- 7-8 points: expression is relatively clear, with occasional minor flaws
- 5-6 points: expression is basically clear, with some areas that are not clear enough
- 0-4 points: expression is not clear enough, with many ambiguities

2. engagement

Evaluation points:
- Whether the opening catches the eye
- Whether the content is engaging
- Whether the pacing is tight
- Whether it makes people willing to continue reading
Scoring criteria (0-10 points):
- 9-10 points: very engaging, fascinating
- 7-8 points: relatively engaging, with highlights
- 5-6 points: Somewhat attractive, but not outstanding enough
- 0-4 points: Insufficient appeal, difficult to spark interest

3. Persuasion

Evaluation points:
- Whether the arguments are sufficient
- Whether the logic is rigorous
- Whether it can effectively influence readers
- Whether it has a call to action
Scoring criteria (0-10 points):
- 9-10 points: Very strong persuasion, rigorous logic
- 7-8 points: Relatively strong persuasion, persuasive
- 5-6 points: Somewhat persuasive, but not sufficient enough
- 0-4 points: Insufficient persuasion, difficult to influence readers

4. Professionalism

Evaluation points:
- Whether the vocabulary used is professional
- Whether the expression is standardized
- Whether there are grammatical errors
- Whether the format is standardized
Scoring criteria (0-10 points):
- 9-10 points: Very professional, standardized and rigorous
- 7-8 points: Relatively professional, with minor flaws
- 5-6 points: Basically professional, with some non-standard aspects
- 0-4 points: Insufficient professionalism, with many problems

5. Creativity

Evaluation points:
- Whether the expression is novel
- Whether there is a unique angle
- Whether there are creative elements
- Whether it is impressive
Scoring criteria (0-10 points):
- 9-10 points: Very creative, highly original
- 7-8 points: Relatively creative, with highlights
- 5-6 points: Somewhat creative, but not outstanding enough
- 0-4 points: Insufficient creativity, relatively bland

6. Readability

Evaluation points:
- Whether sentence length is moderate
- Whether paragraph structure is reasonable
- Whether the wording is easy to understand
- Whether the reading experience is smooth
Scoring criteria (0-10 points):
- 9-10 points: Very good readability, smooth reading experience
- 7-8 points: Relatively good readability, with a few minor issues
- 5-6 points: Average readability, with some areas that are not smooth enough
- 0-4 points: Poor readability, poor reading experience

7. Emotional impact

Evaluation points:
- Whether it can evoke emotional resonance
- Whether the emotional expression is in place
- Whether it is emotionally compelling
- Whether it can leave a deep impression
Scoring criteria (0-10 points):
- 9-10 points: Very strong emotional impact, deep resonance
- 7-8 points: Relatively strong emotional impact, resonant
- 5-6 points: Some emotional impact, but not deep enough
- 0-4 points: Insufficient emotional impact, difficult to evoke resonance

8. Target audience fit

Evaluation points:
- Whether the language style fits the audience's habits
- Whether the content meets the audience's needs
- Whether it is close to the audience's interests
- Whether it can spark audience interaction
Scoring criteria (0-10 points):
- 9-10 points: Highly aligned with the target audience
- 7-8 points: Fairly aligned with the target audience
- 5-6 points: Basically aligned, with room for improvement
- 0-4 points: Not sufficiently aligned with the target audience

Evaluation Process

1. Version Analysis

Analyze the content characteristics of each version
Identify the strengths and weaknesses of each version
Extract the key information of each version

2. Dimension Scoring

Score each version according to the selected evaluation_criteria
Each dimension is scored from 0-10, in increments of 1 point
Calculate the average total score of each version

3. Comparative Analysis

Compare the performance of different versions across dimensions
Identify the strengths and weaknesses of each version
Analyze the differences between versions

4. Comprehensive Evaluation

Comprehensively consider performance across dimensions
Combine target_audience and use_case
Provide version selection recommendations

5. Optimization Suggestions

Provide optimization suggestions for each version
Point out directions for improvement
Provide specific revision suggestions

Version Ranking Method

Comprehensive Score Ranking

Calculate the average total score of each version
Rank from highest to lowest total score
Consider the weight of each dimension (adjusted according to use_case)

Scenario Fit Ranking

Adjust the weight of each dimension according to use_case
Calculate the weighted total score
Rank by weighted total score

For example:

Marketing promotion scenario: engagement and persuasion have higher weight
Information delivery scenario: clarity and readability have higher weight
Brand building scenario: professionalism and target_audience_fit have higher weight

Evaluation Report Structure

1. Overview

Number of versions
Evaluation criteria
Target audience
Use case

2. Detailed Analysis of Each Version

Version A Analysis
- Version Information
- Scores by Dimension
- Strengths and Weaknesses Analysis
- Optimization Suggestions
Version B Analysis
- ...

3. Comparative Analysis

Comparison table of scores for each version
Comparison of performance across dimensions
Analysis of differences between versions

4. Version Ranking

Comprehensive score ranking
Scenario fit ranking
Recommended version

5. Selection Recommendations

Version selection recommendations based on different needs
Best applicable scenarios for each version
Combination optimization suggestions

Quick Mode

When the user input contains complete versions, directly execute using the default parameters (all evaluation criteria, general public audience, automatically determine the scenario based on the content, comparison + scoring evaluation, provide optimization suggestions).

Interaction Question Template (only displayed when the user has not provided the necessary parameters)

我将为您评估这些版本，请确认以下要求：

1. 版本列表（必选）：
   请提供需要评估的版本（2-5个）
   格式示例：
   版本A：（内容）
   版本B：（内容）

2. 评估标准（可多选，可选）：
   [1] 清晰度
   [2] 吸引力
   [3] 说服力
   [4] 专业性
   [5] 创意性
   [6] 可读性
   [7] 情感影响力
   [8] 目标受众契合度

3. 目标受众（可选）：
   [1] 普通大众
   [2] 专业人士
   [3] 年轻群体
   [4] 商务人士
   [5] 学术界
   [6] 自定义（请描述）

4. 使用场景（可选）：
   [1] 营销推广
   [2] 信息传达
   [3] 教育培训
   [4] 娱乐
   [5] 品牌建设
   [6] 自定义

5. 评估方法（可选）：
   [1] 对比评估
   [2] 评分评估
   [3] 对比+评分（推荐）

6. 是否提供优化建议（可选）：
   [1] 提供各版本的优化建议（推荐）
   [2] 不提供

您可以：
- 一次性回复所有信息
- 回复部分参数，我再确认是否需要补充
- 回复"评估"或"好的"，使用推荐设置执行（推荐：全部评估标准，普通大众受众，根据内容自动判断场景，对比+评分，提供优化建议）
- 如果你有其他想法，可在回复的时候一并描述告知

请回复：

Examples

Example 1: Comparative Evaluation of Two Versions

Input：

versions: [
  {"version_name": "版本A", "content": "我们的产品功能强大，操作简单，界面友好。欢迎体验！"},
  {"version_name": "版本B", "content": "这款产品超好用，功能超强，上手超简单，界面超好看！赶紧来试试！"}
]
evaluation_criteria: ["clarity", "engagement", "persuasion", "readability"]
evaluation_method: "both"

Output：

【版本评估报告】

概述：
- 版本数量：2个
- 评估标准：清晰度、吸引力、说服力、可读性
- 目标受众：普通大众
- 使用场景：营销推广

---

各版本详细分析：

【版本A】
内容："我们的产品功能强大，操作简单，界面友好。欢迎体验！"

各维度评分：
- 清晰度：9/10 - 表达非常清晰，无歧义
- 吸引力：6/10 - 表达较为平淡，缺乏吸引力
- 说服力：7/10 - 有说服力，但不够强烈
- 可读性：10/10 - 可读性非常好

总分：8/10

优点：
- 表达清晰简洁
- 语言规范专业
- 可读性强

缺点：
- 吸引力不足，较平淡
- 说服力不够强

优化建议：
1. 增强吸引力：可以使用更生动的表达，如"功能超强"、"超好用"
2. 增强说服力：可以加入具体数据或案例，如"已有10万用户选择"

---

【版本B】
内容："这款产品超好用，功能超强，上手超简单，界面超好看！赶紧来试试！"

各维度评分：
- 清晰度：8/10 - 表达较清晰，但"超"字重复使用略显冗余
- 吸引力：9/10 - 非常吸引人，活泼生动
- 说服力：8/10 - 说服力较强，有号召力
- 可读性：7/10 - 可读性较好，但"超"字重复略显啰嗦

总分：8/10

优点：
- 非常吸引人，活泼生动
- 有强烈的行动号召力
- 符合年轻受众喜好

缺点：
- "超"字重复使用略显冗余
- 专业性略低

优化建议：
1. 减少"超"字的使用，避免重复
2. 可以适当提升专业性，如将"超好用"改为"非常好用"

---

对比分析：

| 维度 | 版本A | 版本B | 胜出 |
|------|-------|-------|------|
| 清晰度 | 9/10 | 8/10 | 版本A |
| 吸引力 | 6/10 | 9/10 | 版本B |
| 说服力 | 7/10 | 8/10 | 版本B |
| 可读性 | 10/10 | 7/10 | 版本A |

综合对比：
- 版本A在清晰度和可读性上更优
- 版本B在吸引力和说服力上更优
- 两个版本各有优势，适合不同场景

---

版本排名：

1. 综合评分排名：
   第1名：版本A（8/10）和版本B（8/10）并列

2. 场景适配排名（营销推广场景，吸引力和说服力权重更高）：
   第1名：版本B
   第2名：版本A

---

选择建议：

1. 如果目标是提升品牌形象和专业性：
   推荐：版本A
   理由：版本A更专业、更规范，适合品牌建设

2. 如果目标是提升转化率和用户点击：
   推荐：版本B
   理由：版本B更具吸引力，有更强的行动号召力

3. 如果目标受众是年轻群体：
   推荐：版本B
   理由：版本B的语言风格更符合年轻受众喜好

4. 如果目标受众是商务人士：
   推荐：版本A
   理由：版本A更专业、更正式

---

组合优化建议：

可以考虑结合两个版本的优点，创建一个优化版本：

优化版本：
"这款产品功能强大，非常好用，操作简单，界面友好。已有10万用户选择，赶紧来体验！"

优化要点：
- 保留版本B的吸引力："这款"、"非常好用"、"赶紧来体验！"
- 保留版本A的专业性："功能强大"、"操作简单"、"界面友好"
- 增加说服力："已有10万用户选择"

Example 2: Multi-Version Evaluation (Marketing Promotion Scenario)

Input：

versions: [
  {"version_name": "版本A", "content": "产品限时特惠，原价299元，现在只需99元！", "version_info": "直接折扣型"},
  {"version_name": "版本B", "content": "价值299元的产品，今天99元带回家！仅限今日！", "version_info": "限时强调型"},
  {"version_name": "版本C", "content": "99元即可拥有价值299元的产品，还赠送VIP会员！", "version_info": "赠送优惠型"}
]
evaluation_criteria: ["clarity", "engagement", "persuasion"]
target_audience: "general"
use_case: "marketing"
evaluation_method: "both"
provide_optimization_suggestions: true

Output：

【版本评估报告】

概述：
- 版本数量：3个
- 评估标准：清晰度、吸引力、说服力
- 目标受众：普通大众
- 使用场景：营销推广

---

各版本详细分析：

【版本A：直接折扣型】
内容："产品限时特惠，原价299元，现在只需99元！"

各维度评分：
- 清晰度：10/10 - 非常清晰，价格信息一目了然
- 吸引力：7/10 - 有吸引力，但略显平淡
- 说服力：8/10 - 说服力较强，价格对比明显

总分：8.3/10

优点：
- 价格信息非常清晰
- 直接展示优惠力度
- 简洁明了

缺点：
- 吸引力不够强，缺乏情感冲击
- 缺少紧迫感

优化建议：
1. 增强吸引力：可以加入情感元素，如"超值"、"必抢"
2. 增强紧迫感：可以强调限时，如"今日特惠"、"仅限今日"

---

【版本B：限时强调型】
内容："价值299元的产品，今天99元带回家！仅限今日！"

各维度评分：
- 清晰度：9/10 - 清晰明了，信息完整
- 吸引力：9/10 - 非常吸引人，情感冲击强
- 说服力：9/10 - 说服力很强，有紧迫感

总分：9/10

优点：
- 非常吸引人，情感冲击强
- 紧迫感强，激发购买冲动
- 表达生动，"带回家"很接地气

缺点：
- 基本无明显缺点

优化建议：
- 已经很优秀，可以保持

---

【版本C：赠送优惠型】
内容："99元即可拥有价值299元的产品，还赠送VIP会员！"

各维度评分：
- 清晰度：8/10 - 较清晰，但信息稍多
- 吸引力：8/10 - 较有吸引力，有赠送亮点
- 说服力：9/10 - 说服力很强，赠送更有吸引力

总分：8.3/10

优点：
- 赠送增加了吸引力
- 性价比感知强
- 说服力强

缺点：
- 信息稍多，清晰度略逊于版本A和B
- 缺少紧迫感

优化建议：
1. 增强紧迫感：可以加上"今日特惠"、"限时"等
2. 提升清晰度：可以简化表达，如"99元=价值299元产品+VIP会员"

---

对比分析：

| 维度 | 版本A | 版本B | 版本C | 胜出 |
|------|-------|-------|-------|------|
| 清晰度 | 10/10 | 9/10 | 8/10 | 版本A |
| 吸引力 | 7/10 | 9/10 | 8/10 | 版本B |
| 说服力 | 8/10 | 9/10 | 9/10 | 版本B&C |

综合对比：
- 版本A在清晰度上最优
- 版本B在吸引力和说服力上最优
- 版本C在说服力上表现优秀，赠送是亮点

---

版本排名：

1. 综合评分排名：
   第1名：版本B（9/10）
   第2名：版本A和版本C（8.3/10）并列

2. 场景适配排名（营销推广场景，吸引力和说服力权重更高）：
   第1名：版本B
   第2名：版本C
   第3名：版本A

---

选择建议：

1. 如果目标是快速转化，强调紧迫感：
   推荐：版本B
   理由：版本B紧迫感最强，吸引力最大，最容易激发购买冲动

2. 如果目标是传递超值感，强调性价比：
   推荐：版本C
   理由：版本C通过赠送，增强了超值感，性价比感知更强

3. 如果目标是清晰传达信息，避免信息过载：
   推荐：版本A
   理由：版本A最清晰简洁，信息最容易理解

---

组合优化建议：

可以考虑结合版本B的紧迫感和版本C的赠送，创建一个终极版本：

终极优化版本：
"价值299元的产品，今天99元带回家！还赠送VIP会员！仅限今日！"

优化要点：
- 保留版本B的紧迫感和生动表达："今天"、"带回家"、"仅限今日！"
- 保留版本C的赠送亮点："还赠送VIP会员！"
- 价格对比清晰：价值299元 vs 99元

Example 3: Scoring evaluation (no optimization suggestions provided)

Input：

versions: [
  {"version_name": "版本A", "content": "人工智能技术正在改变我们的生活。"},
  {"version_name": "版本B", "content": "AI技术正以前所未有的速度改变着我们的生活。"},
  {"version_name": "版本C", "content": "人工智能技术的发展正在深刻影响着我们生活的方方面面。"}
]
evaluation_criteria: ["clarity", "professionalism", "readability"]
evaluation_method: "scoring"
provide_optimization_suggestions: false

Output：

【版本评估报告】

概述：
- 版本数量：3个
- 评估标准：清晰度、专业性、可读性
- 目标受众：普通大众
- 使用场景：信息传达

---

各版本详细分析：

【版本A】
内容："人工智能技术正在改变我们的生活。"

各维度评分：
- 清晰度：10/10 - 表达非常清晰，无歧义
- 专业性：8/10 - 使用规范术语，专业度较高
- 可读性：10/10 - 句子简洁，可读性非常好

总分：9.3/10

优点：
- 表达简洁清晰
- 术语使用规范
- 易于理解

缺点：
- 表达较为平淡，缺乏深度

---

【版本B】
内容："AI技术正以前所未有的速度改变着我们的生活。"

各维度评分：
- 清晰度：8/10 - 表达较清晰，但"前所未有的速度"略显夸张
- 专业性：7/10 - "AI"是缩写，不如"人工智能"规范
- 可读性：9/10 - 句子略长，但可读性较好

总分：8/10

优点：
- 表达较为生动
- 有一定的感染力

缺点：
- 使用缩写"AI"，不够规范
- "前所未有的速度"略显夸张

---

【版本C】
内容："人工智能技术的发展正在深刻影响着我们生活的方方面面。"

各维度评分：
- 清晰度：7/10 - 表达略显冗长，"方方面面"稍显模糊
- 专业性：9/10 - 使用规范术语，表达专业
- 可读性：7/10 - 句子较长，可读性一般

总分：7.7/10

优点：
- 表达专业规范
- "深刻影响"有深度

缺点：
- 句子过长
- "方方面面"略显模糊，影响清晰度

---

版本排名：

综合评分排名：
第1名：版本A（9.3/10）
第2名：版本B（8/10）
第3名：版本C（7.7/10）

---

选择建议：

1. 如果目标是简洁清晰地传达信息：
   推荐：版本A
   理由：版本A最简洁清晰，最易于理解

2. 如果目标是表达生动，有感染力：
   推荐：版本B
   理由：版本B表达较为生动，有一定的感染力

3. 如果目标是专业表达，有深度：
   推荐：版本C
   理由：版本C表达最为专业，有深度

4. 综合推荐：
   版本A的综合表现最优，最推荐使用

Notes

The evaluation is based on AI analysis and is for reference only; the final decision should be made by a human.
In different scenarios, the weights of each evaluation criterion should be adjusted.
The target audience and use scenario have an important impact on the evaluation results; please fill them in accurately.
Optimization suggestions are proposed based on the current version and are for reference only.
The evaluation results may differ from actual A/B test results; it is recommended to make a comprehensive judgment based on actual data.

Parameter Adjustment Tips

After generating content, if you are not satisfied with the current result, you can try adjusting the following parameters:

💡 Parameter adjustment options:

Evaluation Criteria: clarityClarity | engagementAppeal | persuasionPersuasiveness | professionalismProfessionalism | creativityCreativity | readabilityReadability | emotional_impactEmotional Impact | target_audience_fitTarget Audience Fit
Target Audience: generalGeneral Public | professionalProfessionals | youthYoung People | businessBusiness Professionals | academicAcademia
Use Scenario: marketingMarketing Promotion | informationInformation Communication | educationEducation and Training | entertainmentEntertainment | brand_buildingBrand Building
Evaluation Method: comparativeComparative Evaluation | scoringScoring Evaluation | bothComparison + Scoring
Whether to Provide Optimization Suggestions: trueProvide Optimization Suggestions | falseDo Not Provide

Usage: Reply directly, such as "focus on evaluating appeal and persuasiveness" or "change to a marketing promotion scenario"

Relationship with Other Skills

review_text: If a single version needs a comprehensive review, this skill should be used
rewrite_content: If a version needs to be rewritten, this can be used together
check_style_consistency: If the style consistency of multiple versions needs to be checked, this skill should be used
content_imitation: If you want version A to imitate the style of version B, this skill can be used

Configuration File Reference

intent_keywords: ["version evaluation", "version comparison", "A/B testing", "version selection", "multi-version evaluation"]
evaluation_criteria_weights: define the weights of each evaluation criterion in different scenarios in config/evaluation_weights.json

About this skill

Introduction

Function 1: Review Text

Introduction

Use Cases

Input Parameters

Required Parameters

Optional Parameters

Output

Processing Rules

Review Dimensions

1. Grammar check (grammar)

2. Spelling check (spelling)

3. Punctuation check (punctuation)

4. Logic check (logic)

5. Clarity check (clarity)

6. Consistency check (consistency)

7. Factual accuracy check (factual_accuracy)

8. Compliance check (compliance)

Issue severity classification

1. Critical issues (Critical)

2. Important issues (Important)

3. Minor issues (Minor)

Review process

1. Preliminary scan

2. In-depth analysis

3. Issue classification

4. Revision suggestions

5. Overall assessment

Revision suggestion format

Summary format (summary_only)

Detailed format (detailed)

Marked format (with_marked_text)

Highlighting rules

Deleted content

Added content

Modified content

Quick mode

Interactive question template (only displayed when the user does not provide the necessary parameters)

Example

Example 1: Comprehensive proofreading

Example 2: Compliance check

Example 3: Detailed format + highlighting

Notes

Parameter adjustment tips

Relationship with other skills

Configuration file references

Feature 2: Evaluate Versions

Introduction

Use cases

Input parameters

Required parameters

Optional parameters

Output

Processing rules

Detailed explanation of evaluation criteria

1. clarity

2. engagement

3. Persuasion

4. Professionalism

5. Creativity

6. Readability

7. Emotional impact

8. Target audience fit

Evaluation Process

1. Version Analysis

2. Dimension Scoring

3. Comparative Analysis

4. Comprehensive Evaluation

5. Optimization Suggestions

Version Ranking Method

Comprehensive Score Ranking

Scenario Fit Ranking

Evaluation Report Structure

1. Overview

2. Detailed Analysis of Each Version

3. Comparative Analysis

4. Version Ranking

5. Selection Recommendations