30个“AI味“词汇替换全清单：SimpleEnglish一招根治leverage、seamlessly等AI写作陈词滥调

30个AI味词汇替换全清单SimpleEnglish一招根治leverage、seamlessly等AI写作陈词滥调【免费下载链接】SimpleEnglishAgent skill: make LLMs write docs in ASD-STE100 Simplified Technical项目地址: https://gitcode.com/gh_mirrors/si/SimpleEnglishSimpleEnglish 是一个开源 Agent skill它基于航空航天的 ASD-STE100 简化技术英语标准让 LLM 写文档时自动避开 leverage、seamlessly、robust 这类AI味陈词滥调。本文整理出可直接照抄的30 个 AI 写作词汇替换清单并附上真实改写案例和实测数据帮新手快速去除 AI 写作痕迹。为什么 AI 写出来的文章一股AI味 你肯定见过这种句子Leveraging sqlpipes robust architecture, users canseamlesslysynchronize their Postgres tableswith minimal configuration overhead.Its crucialtoensureyour credentials are configured.AI 生成文本有几个标志性毛病堆砌空洞的形容词、同义词轮替check/verify/confirm 混用、模糊的限定词should、might、冗长的从句。SimpleEnglish 的思路很巧妙借用 40 年前航空航天领域的受控语言标准 ASD-STE100。这套标准的初衷是让疲劳的机修工不可能误读一条指令。规则全是可量化的硬规格如句子不超过 20 词而不是写得清楚点这种空话。AI 味陈词滥调正是这套标准的天然反面——AI slop 作为副作用被一并消灭了。核心规则速览规则消灭的对象指令句 ≤20 词、描述句 ≤25 词一口气读不完的长句一个词 一个意思check/verify/confirm 乱用只用简单时态has been updated → we updated主动语态it should be noted that禁止 should/would/may/might含糊其辞的限定词30 个AI味词汇替换全清单直接抄下面是完整替换清单前 26 条来自项目自带的 word-swaps.md后 4 条补充自 SKILL.md 的高频错误表。判断原则如果一个词不携带任何事实直接删除而不是替换。#AI 味写法替换为说明1leverage, utilizeuseAI 最爱的动词直接换成 use2in order toto删掉两个无用的词3prior tobefore正式腔改口语4ensuremake sure that完整、可执行的表述5it is worth noting that删除纯废话6its important to, crucially删除直接陈述事实删掉修饰留事实7simply, just, easily, seamlessly, effortlessly删除零信息量的填充词8robust, powerful, comprehensive, performant删除或给出可测量的指标空洞形容词9functionalityfunction, feature名词堆砌改具体词10enables you to, allows you toyou can主动、简短11is designed to, aims to删除直接说它做什么意图词改动作12facilitatehelp, make possible拉丁词根改平实词13dive into, delve intoread, examineAI 比喻腔14when it comes tofor砍掉 4 个词15in the event thatif法律腔改普通条件句16due to the fact thatbecause一个词就够了17as needed, as necessary写明具体条件模糊条件要落地18and/or二选一或写 X, or Y, or both消除歧义19e.g. / i.e. / etc.for example / that is / 列全拉丁缩写20gracefully handles说清具体行为如重试3次后停止优雅腔改事实21out of the boxby default术语改默认行为22under the hoodinternally厨房腔改技术词23blazingly fast, state-of-the-artfast附数字/删除营销腔改数据24streamlinemake simpler, make faster抽象改具体25plethora, myriadmany大词改小词26addresses the issue, tacklescorrects the fault, removes the error动词具体化27howeverbut高频错误词28thereforethus, as a result高频错误词29performdo正式腔改日常词30since因为because一词多义改明确这份清单的完整机读版可以直接给模型用word-swaps.md。真实案例Before / After 对比下面全部来自 examples/before-after.md 中未经编辑的真实 AI 输出与加载 skill 后的改写。❌ Before原始 AI 输出Oops! Something went wrong while attempting to establish a connection to the database. Please ensure your credentials have been properly configured and try again.✅ After加载 SimpleEnglish 后Connection to the database failed: the password for userappwas not correct. SetDB_PASSWORDto the correct value, then connect again.再看一段故障排查的改写注意 40 词的长句被拆成了 ≤20 词的指令句且条件if都挪到了句子开头❌ BeforeIf youre connecting to a managed database (RDS, Cloud SQL, etc.), confirm the instance allows connections from sqlpipes IP. You can also try increasing the timeout, since a slow network path can trip the default timeout even when the connection eventually succeeds.✅ AfterMake sure that the host that runs sqlpipe can reach the Postgres port. A firewall or security group usually blocks it.If the database is managed (RDS, Cloud SQL), make sure that the instance accepts connections from the IP of sqlpipe.If the network is slow, increasesource.connect_timeout_secondsin the configuration.如何安装 SimpleEnglish3 步上手 零依赖一个文件夹搞定MIT 协议。第 1 步克隆仓库git clone https://gitcode.com/gh_mirrors/si/SimpleEnglish第 2 步接入你的 AI 工具SimpleEnglish 遵循 Agent Skills 开放标准Claude Code、Cursor、VS Code Copilot、OpenAI Codex、Gemini CLI 等约 25 工具都能直接读取 skills/simple-english/ 目录下的技能定义。如果你的工具不支持 skill 机制可以把 prompts/system-prompt.md 粘贴进系统提示词末尾还有一个 60 token 的精简版。第 3 步下达指令让它写技术文档或直接说一句rewrite this with simple-english。两种模式可选模式适用场景行为Pragmatic默认README、错误信息、日常文档应用全部结构规则保留你的领域词汇如 idempotent、webhookStrict需要 STE 合规审计结构规则 完整词汇纪律74.6% 的实测降幅数字是怎么来的 项目内置了确定性正则检查器 ste_lint.py对 7 个 Claude 模型 × 8 个写作任务112 次生成做了同规则对比原始文件全部提交在仓库中可复现模型基线违规/百词加载 skill 后降幅claude-opus-52.130.3285%claude-opus-4-72.280.4282%claude-sonnet-52.670.5380%平均2.310.5874.6%盲评环节不标注标签、双向顺序打分中skill 输出在 56 组对比里赢了 45 组且所有模型的输出 token 都变少了——去掉填充词后文档更短也更准。完整方法与原始数据见 evals/results/RESULTS.md。关键文件导读文件用途skills/simple-english/SKILL.md技能主文件53 条规则、9 大章节 软件示例skills/simple-english/references/word-swaps.mdAI 味词汇替换表本文 30 条清单的源头skills/simple-english/references/checklist.md交付前的自查清单含可搜索的正则模式skills/simple-english/references/use-cases.md错误信息、Runbook、事故报告等长文场景的适配examples/before-after.md真实改写案例集evals/ste_lint.py确定性违规检查器可对任意文本打分小结复制即用的价值就算不安装本文的 30 条替换清单也能立刻提升你的技术写作根因治疗SimpleEnglish 用 40 年验证过的航空航天受控语言标准把写得像 AI变成可检测、可复现的工程问题别用于营销文案这套规则会刻意删掉所有说服力——博客和品牌文案请用你自己的声音把 AI 味留给聊天把清晰留给文档。️【免费下载链接】SimpleEnglishAgent skill: make LLMs write docs in ASD-STE100 Simplified Technical项目地址: https://gitcode.com/gh_mirrors/si/SimpleEnglish创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考