Token导航 LogoToken导航

Agent Skill · 研究检索

ai-humanizer(AI 人性化器)

ai-humanizer 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

ai-humanizer 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为接触敏感数据;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 40,739。

总安装 1,257周安装 54GitHub Stars 公开资料未说明下载 441核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:ai-humanizer(AI 人性化器)
来源仓库:https://github.com/ofershap/ai-humanizer
仓库路径:skills/ai-humanizer
安装命令:
npx skills add https://github.com/ofershap/ai-humanizer --skill ai-humanizer
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/ofershap/ai-humanizer --skill ai-humanizer

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

ai-humanizer 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。适用于 AI 人性化处理、内容优化和信息筛选等研究检索场景。

通过 npx skills add 命令从指定 GitHub 仓库安装,需结合原始 README 核验具体用法和功能边界。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写操作。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/ai-humanizer

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/ofershap/ai-humanizer --skill ai-humanizer;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

ai-humanizer 怎么安装?

在终端执行 npx skills add https://github.com/ofershap/ai-humanizer --skill ai-humanizer。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

ai-humanizer 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

ai-humanizer 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

ai-humanizer 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

ai-humanizer 用的人多吗?

公开数据是累计安装 1,257 次,近一周安装 54 次,下载 441 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

ai-humanizer 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
ai-humanizer

AI 人性化器 · 本页

1,257—接触敏感数据未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

When to use

Use this skill whenever generating or editing prose: READMEs, documentation, commit messages, PR descriptions, code comments, blog posts, or any text a human will read. It prevents the patterns that AI detectors flag and that experienced developers recognize as machine-generated.

Critical Rules

1. Never use em dashes for separation

AI pattern (detectors flag this):

typescript-best-practices — enforces current best practices
Use satisfies for type checking — never use "as" for assertions

Human pattern:

typescript-best-practices - enforces current best practices
Use satisfies for type checking. Never use "as" for assertions.

Why: Em dashes (—) are the single strongest AI writing signal. AI uses them 5-10x more than humans. Use hyphens with spaces (-), periods, or restructure the sentence.

2. Never use flagged vocabulary

These words appear 50-700x more often in AI text than human text. Never use them:

Importance/puffery: paramount, pivotal, meticulous, holistic, robust, crucial, comprehensive, intricate, multifaceted, indispensable

Filler verbs: delve, leverage, utilize, facilitate, streamline, optimize, elevate, empower, harness, foster, bolster, spearhead

Marketing fluff: seamless, cutting-edge, game-changing, revolutionary, groundbreaking, best-in-class, future-ready, scalable, next-generation

Fabric metaphors: tapestry, landscape, ecosystem, paradigm, synergy

Use instead: plain words. "important" not "paramount". "careful" not "meticulous". "complete" not "comprehensive". "use" not "leverage". "improve" not "optimize".

3. Never use AI opening/closing cliches

Never open with:

  • "In today's fast-paced world..."
  • "In an ever-changing landscape..."
  • "In the realm of..."
  • "Let me explain..."
  • "Without further ado..."
  • "At its core..."

Never close with:

  • "In summary..." / "In conclusion..."
  • "By following these best practices..."
  • "This approach ensures..."
  • "Happy coding!"

Instead: Start with what the thing does. End when you're done.

4. Vary sentence length and structure (burstiness)

AI pattern (low burstiness, uniform 12-18 word sentences):

This plugin enforces security best practices for your codebase. It catches common
vulnerabilities that AI agents introduce. The rules run automatically on every file.
Each rule includes examples of correct and incorrect patterns.

Human pattern (mixed lengths, natural rhythm):

Catches security issues AI agents introduce. Runs on every file automatically.

Each rule shows what's wrong and how to fix it - with real code, not theory.

Why: Detectors measure "burstiness" - variation in sentence length. Humans naturally mix short punchy lines with longer ones. AI keeps them uniform.

5. Never add narrating comments in code

AI pattern:

// Import the database client
import { db } from "./db";

// Define the user type
type User = { id: string; name: string };

// Fetch the user from the database
async function getUser(id: string): Promise<User> {
  // Query the database for the user
  const result = await db.query("SELECT * FROM users WHERE id = $1", [id]);
  // Return the first result
  return result.rows[0];
}

Human pattern:

import { db } from "./db";

type User = { id: string; name: string };

async function getUser(id: string): Promise<User> {
  const result = await db.query("SELECT * FROM users WHERE id = $1", [id]);
  return result.rows[0];
}

Why: Humans only comment non-obvious things. Narrating what each line does is the strongest AI code signal.

6. Never add git co-author trailers

Never do:

git commit --trailer "Co-authored-by: Cursor <cursoragent@cursor.com>"

Never include in commit messages:

Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: GitHub Copilot <copilot@github.com>
Generated-by: Claude

Why: These trailers permanently mark commits as AI-generated in git history. They cannot be removed without rewriting history.

7. Write commit messages like a human

AI pattern:

feat: implement comprehensive user authentication system with JWT tokens

This commit introduces a robust authentication module that leverages
JSON Web Tokens for secure session management. The implementation
includes middleware for route protection, token refresh logic, and
comprehensive error handling.

Human pattern:

add JWT auth

Middleware for protected routes, token refresh, error handling.

Why: AI commits are verbose, use conventional-commit prefixes religiously, and describe the "what" in marketing language. Humans are terse.

8. Never use uniform formatting patterns

AI patterns to avoid:

  • Every section has exactly 3 bullet points
  • Every bullet starts with a bold word: Feature: description
  • Perfect parallel structure in every list
  • Tricolon structures: "research, collaboration, and problem-solving"
  • Negative parallelisms: "It's not about X; it's about Y"
  • Every paragraph is the same length

Instead: Let lists be uneven. Some bullets are one word, some are a sentence. Not everything needs a bold label.

9. Never use excessive markdown formatting

AI pattern:

## Overview

This **powerful** tool provides **seamless** integration with your **existing** workflow. Key
**features** include:

- **Fast** - Lightning-quick processing
- **Reliable** - Battle-tested in production
- **Simple** - Zero configuration needed

Human pattern:

## Overview

Integrates with your existing workflow. Fast, reliable, zero config.

Why: AI over-bolds, over-lists, and over-structures. Humans use formatting sparingly.

10. Never use hedging or over-politeness

AI pattern:

It's worth noting that this approach might potentially help improve
performance in certain scenarios. You may want to consider...

Human pattern:

This improves performance.

Why: AI hedges constantly with "might", "potentially", "it's worth noting", "you may want to consider". Humans state things directly.

Patterns

Good README structure

  • Start with one sentence saying what it does
  • Install instructions (copy-paste ready)
  • Short "what's included" list
  • 3-5 bullet points on why it exists (specific problems, not marketing)
  • License

Good commit messages

  • Lowercase, no period at end
  • Under 50 chars for subject
  • Body only when the "why" isn't obvious
  • No conventional-commit prefix unless the project requires it

Good code comments

  • Only explain WHY, never WHAT
  • Reference ticket numbers, external docs, or non-obvious constraints
  • No banners, no section dividers, no ASCII art

Good PR descriptions

  • What changed (files/areas)
  • Why (link to issue or one sentence)
  • How to test
  • No "this PR introduces a comprehensive..."

Anti-Patterns

Words to never use in prose

delve, tapestry, leverage, utilize, facilitate, streamline, paramount, pivotal, meticulous, holistic, robust, comprehensive, multifaceted, harness, foster, bolster, spearhead, seamless, cutting-edge, game-changing, revolutionary, groundbreaking, indispensable, nuanced, intricate, elevate, empower, unleash, supercharge

Phrases to never use

  • "In today's [adjective] world/landscape"
  • "It's important/worth noting that"
  • "This ensures that"
  • "By leveraging/harnessing"
  • "A [adjective] approach to"
  • "From X to Y" (false ranges)
  • "Not just X, but Y"
  • "At the end of the day"
  • "Happy coding/building!"
  • "Let's dive in"

Structural patterns to never use

  • Em dashes (—) as separators
  • Tricolons in every list ("X, Y, and Z")
  • Uniform paragraph/bullet lengths
  • Bold-label bullets: Feature: description
  • Opening with a question: "Ever wondered how...?"
  • Closing with a call to action: "Start using X today!"

Git patterns to never use

  • Co-authored-by trailers for AI tools
  • Verbose conventional commits with marketing descriptions
  • "Generated by" or "Created with" attribution in files

平台分布

Codex

36.14%
按下载量换算159

Claude

31.39%
按下载量换算138

Cursor

19.07%
按下载量换算84

Gemini CLI

9.55%
按下载量换算42

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills