proofreading
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseProofreading
文本校对
This skill enables an AI agent to systematically proofread text, catching errors in grammar, spelling, punctuation, style, and consistency. The agent applies a structured multi-pass review process, supports standard style guides (AP, Chicago, APA, or custom house styles), and provides readability scoring alongside tone analysis. Output includes both the corrected text and an annotated summary of every change made.
该技能可让AI Agent对文本进行系统性校对,捕捉语法、拼写、标点、风格和一致性方面的错误。Agent采用结构化的多轮审核流程,支持标准风格指南(AP、Chicago、APA或自定义内部风格),并在进行语气分析的同时提供可读性评分。输出内容包括修正后的文本以及所有修改的注释摘要。
Workflow
工作流程
-
Grammar Check Scan the text for grammatical errors including subject-verb agreement, incorrect tense usage, dangling modifiers, sentence fragments, run-on sentences, and misplaced clauses. Fix each error and record the original phrasing alongside the correction. Pay special attention to complex sentences where multiple clauses may introduce ambiguity.
-
Spelling and Word Choice Identify misspelled words, commonly confused homophones (e.g., "affect" vs. "effect," "their" vs. "there"), and incorrect word forms. Verify proper nouns and domain-specific terminology against the provided context or glossary. Flag any words that are spelled correctly but likely used in the wrong context.
-
Punctuation and Mechanics Review comma usage, semicolons, colons, em dashes, en dashes, hyphens, quotation marks, and apostrophes. Apply the rules of the specified style guide — for example, the Oxford comma for Chicago style but not for AP style. Correct misuse of ellipses, parentheses, and bracket nesting.
-
Style and Consistency Enforce consistent formatting throughout the document: heading capitalization (title case vs. sentence case), number formatting (numerals vs. spelled-out), abbreviation usage (define on first use), and list punctuation. Check for passive voice overuse and recommend active alternatives where the meaning would be clearer. Ensure consistent use of terminology — if "user interface" appears in one place and "UI" elsewhere, unify them.
-
Clarity and Readability Evaluate sentence length, paragraph structure, and overall readability. Flag sentences longer than 30 words for possible splitting. Identify jargon or overly complex vocabulary that could be simplified without losing precision. Calculate a readability score (Flesch-Kincaid or equivalent) and report it. Suggest structural improvements such as adding transition sentences or breaking up dense paragraphs.
-
Tone Analysis and Final Review Assess the overall tone (formal, conversational, academic, persuasive) and verify it matches the user's stated intent. Flag any jarring tonal shifts — for instance, a casual phrase in an otherwise formal report. Perform a final read-through to ensure all corrections integrate smoothly and no new errors were introduced during editing.
-
语法检查 扫描文本中的语法错误,包括主谓一致、时态误用、悬垂修饰语、句子碎片、连写句和错位从句。修正每个错误并记录原文表述与修正内容。特别注意包含多个从句、可能引发歧义的复杂句子。
-
拼写与用词选择 识别拼写错误的单词、易混淆的同音异义词(如"affect"与"effect"、"their"与"there")以及不正确的单词形式。根据提供的上下文或术语表核实专有名词和领域特定术语。标记拼写正确但可能使用语境错误的单词。
-
标点与格式规范 检查逗号、分号、冒号、破折号、连字符、引号和撇号的使用。遵循指定风格指南的规则——例如,Chicago风格要求牛津逗号,而AP风格则不要求。修正省略号、括号和嵌套括号的误用。
-
风格与一致性 确保文档格式一致:标题大小写(标题式大小写 vs 句子式大小写)、数字格式(阿拉伯数字 vs 拼写形式)、缩写使用(首次出现时定义)和列表标点。检查被动语态过度使用的情况,并在表意更清晰的情况下建议使用主动语态替代。确保术语使用一致——如果一处出现"user interface",另一处出现"UI",则统一术语。
-
清晰度与可读性 评估句子长度、段落结构和整体可读性。标记超过30个单词的句子,建议拆分。识别可简化且不丢失精准度的行话或过于复杂的词汇。计算可读性评分(Flesch-Kincaid或等效指标)并报告。建议结构改进,例如添加过渡句或拆分密集段落。
-
语气分析与最终审核 评估整体语气(正式、口语化、学术性、说服性)并验证其是否符合用户声明的意图。标记任何突兀的语气转变——例如,正式报告中出现的随意短语。进行最终通读,确保所有修正整合流畅,且编辑过程中未引入新错误。
Usage
使用方法
Provide the text to proofread along with any relevant context. Optionally specify:
- Style guide — AP, Chicago, APA, MLA, or a custom style reference
- Audience — academic, professional, general public, technical
- Tone — formal, informal, conversational, neutral
- Focus areas — grammar only, style only, full proofread
- Output format — corrected text only, corrected text with tracked changes, or annotated list of changes
Example prompt: "Proofread the following research abstract for APA style. Flag any passive voice usage and provide a readability score."
提供需要校对的文本及相关上下文。可选择性指定:
- 风格指南——AP、Chicago、APA、MLA或自定义风格参考
- 受众——学术、专业、普通大众、技术领域
- 语气——正式、非正式、口语化、中立
- 重点领域——仅语法、仅风格、全面校对
- 输出格式——仅修正后文本、带修订痕迹的修正后文本或修改注释列表
示例提示:"按照APA风格校对以下研究摘要。标记所有被动语态用法并提供可读性评分。"
Examples
示例
Example 1: Proofreading an Academic Paragraph
示例1:学术段落校对
Input:
The study's participants was selected from a pool of undergraduate students, who where enrolled in a introductory psychology course. Data was collected over a 6 week period and analysed using a mix-methods approach. The researchers founded that participants that recieved the intervention showed signficant improvement, however those in the control group did not.
Corrected Output:
The study's participants were selected from a pool of undergraduate students who were enrolled in an introductory psychology course. Data were collected over a six-week period and analyzed using a mixed-methods approach. The researchers found that participants who received the intervention showed significant improvement; however, those in the control group did not.
Changes Made:
| # | Original | Correction | Reason |
|---|---|---|---|
| 1 | "participants was" | "participants were" | Subject-verb agreement |
| 2 | "who where" | "who were" | Homophone error |
| 3 | "a introductory" | "an introductory" | Article agreement (vowel sound) |
| 4 | "Data was" | "Data were" | "Data" is plural in academic writing (APA) |
| 5 | "6 week" | "six-week" | Spell out numbers under 10; hyphenate compound modifier |
| 6 | "analysed" | "analyzed" | American English spelling (APA default) |
| 7 | "mix-methods" | "mixed-methods" | Incorrect adjective form |
| 8 | "founded" | "found" | Wrong verb form |
| 9 | "participants that" | "participants who" | Use "who" for people |
| 10 | "recieved" | "received" | Spelling error |
| 11 | "signficant" | "significant" | Spelling error |
| 12 | "improvement, however" | "improvement; however," | Semicolon before conjunctive adverb |
Readability: Flesch-Kincaid Grade Level 14.2 (appropriate for academic audience).
输入:
The study's participants was selected from a pool of undergraduate students, who where enrolled in a introductory psychology course. Data was collected over a 6 week period and analysed using a mix-methods approach. The researchers founded that participants that recieved the intervention showed signficant improvement, however those in the control group did not.
修正后输出:
The study's participants were selected from a pool of undergraduate students who were enrolled in an introductory psychology course. Data were collected over a six-week period and analyzed using a mixed-methods approach. The researchers found that participants who received the intervention showed significant improvement; however, those in the control group did not.
修改内容:
| 序号 | 原文 | 修正内容 | 修改原因 |
|---|---|---|---|
| 1 | "participants was" | "participants were" | 主谓一致 |
| 2 | "who where" | "who were" | 同音异义词错误 |
| 3 | "a introductory" | "an introductory" | 冠词搭配(元音开头) |
| 4 | "Data was" | "Data were" | 学术写作中"Data"为复数(APA风格) |
| 5 | "6 week" | "six-week" | 10以下数字拼写;复合修饰语加连字符 |
| 6 | "analysed" | "analyzed" | 美式英语拼写(APA默认) |
| 7 | "mix-methods" | "mixed-methods" | 形容词形式错误 |
| 8 | "founded" | "found" | 动词形式错误 |
| 9 | "participants that" | "participants who" | 指代人使用"who" |
| 10 | "recieved" | "received" | 拼写错误 |
| 11 | "signficant" | "significant" | 拼写错误 |
| 12 | "improvement, however" | "improvement; however," | 连接副词前使用分号 |
可读性: Flesch-Kincaid年级水平14.2(适合学术受众)。
Example 2: Proofreading Technical Documentation
示例2:技术文档校对
Input:
Configration
To setup the SDK, first install the dependancies using npm. Runin your terminal. After installing you need to initialise the client by callingnpm install @acme/sdkconstructor and passing you're API key. Make sure the api key is stored securly and is not hard-coded in your source code, we reccommend using environment variables.AcmeClient()
Corrected Output:
Configuration
To set up the SDK, first install the dependencies using npm. Runin your terminal. After installing, you need to initialize the client by calling thenpm install @acme/sdkconstructor and passing your API key. Make sure the API key is stored securely and is not hard-coded in your source code. We recommend using environment variables.AcmeClient()
Changes Made:
| # | Original | Correction | Reason |
|---|---|---|---|
| 1 | "Configration" | "Configuration" | Spelling error in heading |
| 2 | "To setup" | "To set up" | "Set up" is a verb phrase; "setup" is a noun/adjective |
| 3 | "dependancies" | "dependencies" | Spelling error |
| 4 | "After installing you" | "After installing, you" | Missing comma after introductory phrase |
| 5 | "initialise" | "initialize" | American English spelling for consistency |
| 6 | "calling | "calling the | Missing article |
| 7 | "you're" | "your" | Homophone error ("you're" = "you are") |
| 8 | "api key" | "API key" | Acronym should be capitalized |
| 9 | "securly" | "securely" | Spelling error |
| 10 | "source code, we" | "source code. We" | Comma splice — use a period to separate independent clauses |
| 11 | "reccommend" | "recommend" | Spelling error |
输入:
Configration
To setup the SDK, first install the dependancies using npm. Runin your terminal. After installing you need to initialise the client by callingnpm install @acme/sdkconstructor and passing you're API key. Make sure the api key is stored securly and is not hard-coded in your source code, we reccommend using environment variables.AcmeClient()
修正后输出:
Configuration
To set up the SDK, first install the dependencies using npm. Runin your terminal. After installing, you need to initialize the client by calling thenpm install @acme/sdkconstructor and passing your API key. Make sure the API key is stored securely and is not hard-coded in your source code. We recommend using environment variables.AcmeClient()
修改内容:
| 序号 | 原文 | 修正内容 | 修改原因 |
|---|---|---|---|
| 1 | "Configration" | "Configuration" | 标题拼写错误 |
| 2 | "To setup" | "To set up" | "Set up"是动词短语;"setup"是名词/形容词 |
| 3 | "dependancies" | "dependencies" | 拼写错误 |
| 4 | "After installing you" | "After installing, you" | 介绍性短语后缺少逗号 |
| 5 | "initialise" | "initialize" | 美式英语拼写以保持一致性 |
| 6 | "calling | "calling the | 缺少冠词 |
| 7 | "you're" | "your" | 同音异义词错误("you're" = "you are") |
| 8 | "api key" | "API key" | 首字母缩写应大写 |
| 9 | "securly" | "securely" | 拼写错误 |
| 10 | "source code, we" | "source code. We" | 逗号拼接错误——使用句号分隔独立从句 |
| 11 | "reccommend" | "recommend" | 拼写错误 |
Best Practices
最佳实践
- Use a multi-pass approach. Never try to catch all error types in a single read. Separate passes for grammar, spelling, punctuation, and style yield more thorough results.
- Preserve the author's voice. Correct errors without rewriting the text in a different style. The goal is to polish, not to replace the author's expression.
- Show your work. Always provide an annotated list of changes so the author can review, learn from corrections, and accept or reject individual edits.
- Apply the correct style guide consistently. AP and Chicago have different rules for commas, numbers, and capitalization. Confirm which guide applies before starting.
- Flag subjective suggestions separately. Distinguish between objective errors (misspellings, grammar mistakes) and subjective suggestions (word choice, sentence restructuring) so the author can prioritize.
- Consider the audience. A technical document for developers can tolerate jargon that would be inappropriate in a public-facing blog post. Calibrate simplification suggestions accordingly.
- 采用多轮审核方法。切勿尝试单次通读捕捉所有类型的错误。分别针对语法、拼写、标点和风格进行审核,结果会更全面。
- 保留作者的语气。修正错误时不要以不同风格重写文本。目标是润色,而非替代作者的表达。
- 展示修改过程。始终提供修改注释列表,以便作者审核、从修正中学习,并接受或拒绝个别编辑。
- 始终遵循正确的风格指南。AP和Chicago在逗号、数字和大写规则上存在差异。开始前确认适用的指南。
- 单独标记主观建议。区分客观错误(拼写错误、语法错误)和主观建议(用词选择、句子重组),以便作者优先处理。
- 考虑受众。面向开发者的技术文档可以容忍行话,而这些行话在面向公众的博客文章中并不合适。据此调整简化建议。
Edge Cases
边缘情况
- Intentional style deviations: Creative writing, dialogue, and marketing copy may intentionally break grammar rules for effect. Ask for clarification before "correcting" sentence fragments or colloquialisms that appear purposeful.
- Mixed languages: Documents containing code snippets, foreign phrases, or brand names in another language should have those sections excluded from spelling checks.
- Regional English variants: Distinguish between American, British, Australian, and other English variants. "Colour" is not a misspelling in British English; "analyze" is not a misspelling in American English.
- Highly technical or domain-specific text: Medical, legal, and scientific texts contain specialized terminology that may trigger false positives. Cross-reference flagged terms against domain glossaries before marking them as errors.
- Very short text: For single sentences or headlines, readability scoring may not be meaningful. Focus on correctness and clarity instead.
- 故意偏离风格:创意写作、对话和营销文案可能为了效果而故意违反语法规则。在“修正”看似有意的句子碎片或口语表达前,先请求澄清。
- 混合语言:包含代码片段、外语短语或其他语言品牌名称的文档,这些部分应排除在拼写检查之外。
- 英语区域变体:区分美式、英式、澳式及其他英语变体。在英式英语中,“Colour”不是拼写错误;在美式英语中,“analyze”不是拼写错误。
- 高度技术化或领域特定文本:医学、法律和科学文本包含专业术语,可能会触发误报。在标记为错误前,对照领域术语表交叉验证标记的术语。
- 极短文本:对于单句或标题,可读性评分可能没有意义。重点关注正确性和清晰度。