← 返回 Skills 市场
terryso

de-ai-flavor

作者 NEE · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ 安全检测通过
73
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install de-ai-flavor
功能描述
检查并优化中文文章的AI翻译腔,让文本更自然
使用说明 (SKILL.md)

去AI味

检查中文文章中的 AI 翻译腔,给出修改建议并直接优化。

输入

{ARGUMENT} 是要检查的文件路径。如果为空,询问用户提供文件路径。

检测规则

逐段扫描文章,检测以下 4 类 AI 翻译腔:

规则 1:物理动作动词形容思维

用物理动作来描述抽象思维过程,是英文直译的典型痕迹。

常见问题词: 接住、击穿、拆解、锋利、打通、抓住、落地、深耕、赋能、抓手、沉淀、闭环、拉齐、对齐(作动词用时)、链路、痛点、颗粒度

修改方式: 换成直接的中文表达,或删掉。

  • "拆解这个问题" → "分析这个问题"
  • "打通全链路" → "完成全部流程"
  • "赋能开发者" → "帮开发者"

规则 2:形容词预判 + 冒号

先给一个主观评价,再用冒号展开内容。AI 特别爱用这种结构。

典型模式:

  • "X 很简单:"
  • "X 有几个亮点:"
  • "X 最让人印象深刻的是:"
  • "X 的核心理念是:"
  • "X 清晰分层:"

修改方式: 删掉评价部分,直接展开内容。

  • "SDK 的架构清晰分层:" → "SDK 的架构分了几层:"
  • "就这么简单。prompt() 是阻塞式的" → "prompt() 是阻塞式的"
  • "这是最让人印象深刻的特性之一" → 直接说内容

规则 3:抽象名词主语 + 形容词结论

用抽象名词做主语,配上一个形容词结论,看起来高级但空洞。

典型模式:

  • "X 的设计哲学不是 A 而是 B"
  • "X 为 Y 生态带来了一套完整的 Z"
  • "X 提供了一套生产级别的 Y"
  • "这些是在构建 Z 时必不可少的基础设施"

修改方式: 换成具体描述,或直接说结论。

  • "它的设计哲学不是做一个简单的封装,而是提供一套生产级别的运行时" → "它提供的不只是 LLM API 的一层封装"
  • "值得一试" → "可以 clone 下来跑跑看"(给具体动作)

规则 4:不必要的英文混入

中英文夹杂在技术文章中有时必要,但有些混入是多余的。

判断标准: 如果有对应的中文名称且读者普遍熟悉,就不需要英文。

  • "Interface" → "接口"(如果上下文都是中文)
  • "Leverage" → 直接用中文

注意: 技术术语保留英文是正常的(API、SDK、LLM、Agent、MCP 等),不要过度修改。

执行流程

  1. 读取文件:读取 {ARGUMENT} 指定的文件
  2. 逐段扫描:按上面 4 条规则逐段检查,记录所有问题
  3. 汇报问题:列出发现的问题,按规则分类,每条给出原文和修改建议
  4. 应用修改:直接修改文件
  5. 显示 diff:用 git diff 展示改动,让用户确认

输出格式

先汇报问题,格式如下:

规则 X:\x3C规则名称>
- 第 N 段:"原文片段" → 建议改为:"修改后"

然后直接应用修改,最后显示 diff。

注意事项

  • 只修改有问题的句子,不要改写没有 AI 味的正常内容
  • 技术术语的英文保留不动
  • 保持文章原有的结构和排版
  • 如果文章几乎没有 AI 味,直接告诉用户"这篇文章没什么 AI 味,不需要改"
安全使用建议
This skill is coherent for editing Chinese Markdown to remove AI-sounding phrasing, but note: 1) it reads and writes files in-place — back up the file or run it on a copy if you want to avoid accidental overwrites; 2) the SKILL.md says it will show a git diff, but it does not declare git as required — ensure the file is in a git repo or that git is installed, otherwise the diff step may fail or the edit flow may be different; 3) the instructions apply changes before displaying the diff (ambiguous), so confirm with the developer or test on a copy first; 4) review suggested edits carefully — automated rephrasing can change nuance or technical meaning. If you need safer behavior, ask the author to change the flow to 'show proposed edits and ask for confirmation before modifying files.'
功能分析
Type: OpenClaw Skill Name: de-ai-flavor Version: 1.0.0 The skill bundle is designed to identify and remove 'AI-generated' linguistic patterns and corporate jargon from Chinese Markdown files. The instructions in SKILL.md guide the agent to perform text analysis, file modification, and display changes via 'git diff', all of which are consistent with the stated purpose and show no signs of malicious intent or data exfiltration.
能力评估
Purpose & Capability
The name/description (remove 'AI' translationese from Chinese text) matches the instructions: the skill reads a Markdown file, detects translationese patterns, suggests edits, and writes modifications. It does not request unrelated credentials or system access.
Instruction Scope
The SKILL.md instructs the agent to read an arbitrary file path and to '直接修改文件' (modify the file in-place) and then '显示 diff' using git diff. That means the skill will perform filesystem writes without an explicit confirmation step before applying changes (the doc shows diff after applying), and it assumes git is available although no binaries are declared. These behaviors are coherent with a file-editing tool but can lead to unexpected overwrites if the user expects a preview-before-write.
Install Mechanism
No install spec (instruction-only). No code is installed or downloaded, so there's no installation risk from external URLs or packages.
Credentials
The skill requests no environment variables, credentials, or config paths — proportional to a local file-editing tool. It does not attempt to exfiltrate data or call external services in the instructions.
Persistence & Privilege
always is false and the skill does not request persistent/system-wide privileges. Autonomous invocation is allowed (platform default) but is not combined with other concerning privileges or credential access.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install de-ai-flavor
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /de-ai-flavor 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
- Initial release of de-ai-flavor: a tool to detect and optimize AI translation style in Chinese technical articles. - Scans Markdown files for four common types of unnatural AI-originated expressions and suggests natural alternatives. - Applies modifications directly to the file and displays a git diff for user review. - Preserves technical terminology in English when appropriate. - Clearly lists detected issues by rule, showing both original text and suggested edits. - Reports if the article does not have significant AI translation style.
元数据
Slug de-ai-flavor
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

de-ai-flavor 是什么?

检查并优化中文文章的AI翻译腔,让文本更自然. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 73 次。

如何安装 de-ai-flavor?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install de-ai-flavor」即可一键安装,无需额外配置。

de-ai-flavor 是免费的吗?

是的,de-ai-flavor 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

de-ai-flavor 支持哪些平台?

de-ai-flavor 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 de-ai-flavor?

由 NEE(@terryso)开发并维护,当前版本 v1.0.0。

💬 留言讨论