Token导航 LogoToken导航

Agent Skill · 研究检索

human-writer(人类作家)

human-writer 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。

human-writer 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,权限为需要联网;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 65,208。

总安装 285周安装 12GitHub Stars 1下载 411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:human-writer(人类作家)
来源仓库:https://github.com/vassovass/scl-v3
仓库路径:skills/human-writer
安装命令:
npx skills add https://github.com/vassovass/scl-v3 --skill human-writer
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/vassovass/scl-v3 --skill human-writer

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。

human-writer 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词、任务场景或来源线索快速定位候选结果。它可辅助从文档中提取内容、匹配字段或过滤无关信息,提升信息获取效率。

通过 npx skills add 命令从指定仓库安装,具体用法需结合 README 进一步确认。安装前建议检查权限范围和维护状态,避免触发不必要的联网或文件操作。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/human-writer

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/vassovass/scl-v3 --skill human-writer;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

    它需要需要联网,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

human-writer 怎么安装?

在终端执行 npx skills add https://github.com/vassovass/scl-v3 --skill human-writer。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

human-writer 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。

human-writer 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

human-writer 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。

human-writer 用的人多吗?

公开数据是累计安装 285 次,近一周安装 12 次,下载 411 次,GitHub 1 星。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。

human-writer 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

  • 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

  • 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
human-writer

人类作家 · 本页

2851需要联网未明确通过

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

Human Writer

Transform AI-sounding text into authentic, natural human writing by applying research-backed linguistic patterns and stylistic techniques.

When to Use This Skill

Use this skill when:

  • Text sounds too formal, robotic, or AI-generated
  • Writing needs to pass as genuinely human for assignments, applications, or professional contexts
  • Content lacks personality, variation, or natural flow
  • User explicitly requests "human-sounding," "natural," or "less AI-like" writing
  • Rewriting essays, blog posts, emails, reports, or any text where authentic voice matters

Core Approach

Apply these transformations in sequence:

1. Sentence Variation (Burstiness)

Human writing naturally varies sentence length dramatically. AI defaults to medium-length sentences.

Apply:

  • Mix very short sentences (3-7 words) with longer complex ones (20-35 words)
  • Use occasional fragments for emphasis
  • Vary sentence starters - avoid repetitive patterns
  • Break up dense paragraphs with punchy statements

Example transformation:

  • AI: "The research demonstrates that machine learning algorithms can process vast amounts of data efficiently. This capability enables organizations to make better decisions. The technology continues to evolve rapidly."
  • Human: "Machine learning processes massive datasets fast. That's the core capability. Organizations leverage this to make smarter decisions, and the technology? It keeps evolving at breakneck speed."

2. Remove AI Tell-Tale Words and Phrases

AI models overuse specific words that humans rarely employ. See references/ai-tells.md for comprehensive list.

Common offenders to reduce/remove:

  • Overused: delve, tapestry, camaraderie, palpable, intricate, unease, robust, crucial, vital, realm, landscape (metaphorical)
  • Filler transitions: Moreover, Furthermore, Additionally, In addition, Thus, Hence, Therefore (use sparingly)
  • Overly formal: Utilize (use "use"), Commence (use "start"), Facilitate (use "help")

Replace with:

  • Natural connectors: And, But, So, Plus, Also, Besides
  • Casual transitions: Anyway, Basically, Essentially, Look
  • Concrete verbs: Make, do, fix, build, create, help

3. Add Contractions and Informal Language

AI defaults to formal constructions. Humans use contractions freely in most contexts except highly formal academic/legal writing.

Apply liberally:

  • Don't, can't, won't, it's, I'm, we're, they're, you're
  • Wanna, gonna (in very casual contexts only)
  • Colloquialisms appropriate to context

BUT avoid in:

  • Academic research papers
  • Legal documents
  • Formal business proposals
  • When user specifically requests formal tone

4. Vary Sentence Structure

AI overuses certain grammatical patterns. Humans naturally vary structure.

AI overuses:

  • Present participial clauses: "Walking down the street, seeing the crowd..." (use 50% less)
  • Nominalizations: "implementation" → "implementing," "utilization" → "using"
  • Agentless passive voice

Add more:

  • Simple active voice: "The team built it" not "It was built by the team"
  • Questions mid-paragraph: "But does that really work?"
  • Direct address: "You know what I mean?"

5. Inject Personal Voice and Imperfection

Humans think aloud, hesitate, correct themselves, and show personality.

Add strategically:

  • Hedge words: kind of, sort of, pretty much, a bit, somewhat (don't overdo)
  • Thinking markers: I think, I feel, seems like, looks like
  • Self-correction: "Well, actually..." "Or rather..." "I mean..."
  • Qualifiers: mostly, often, usually, typically
  • Minor redundancy: "completely finished," "end result"

Example:

  • AI: "The solution effectively addresses the problem."
  • Human: "This solution pretty much solves the problem, though there's probably room for improvement."

6. Use Natural Collocations and Idioms

AI sometimes creates unnatural word pairings. Humans use familiar phrases.

Check for natural pairings:

  • Make the bed (not do the bed)
  • Do the dishes (not make the dishes)
  • Strong coffee (not powerful coffee)
  • Heavy rain (not strong rain)

Add appropriate idioms:

  • "Hit the ground running"
  • "Cut to the chase"
  • "Get the ball rolling"
  • Context-appropriate only - don't force them

7. Create Contextual Depth

Humans draw from experience and add specific details. AI stays generic.

Add when appropriate:

  • Specific examples instead of generalizations
  • Real-world references (but avoid fabricating facts)
  • Analogies and comparisons
  • Personal observations or hedged experiences

Example:

  • AI: "Many people enjoy coffee in the morning."
  • Human: "Most people I know can't function without coffee before 9am. Myself included."

Progressive Application

For different output requirements:

Light humanization (professional contexts):

  • Apply sentence variation
  • Add occasional contractions
  • Remove most AI-tells
  • Keep sophisticated vocabulary

Medium humanization (blog posts, emails):

  • Full sentence variation including fragments
  • Generous contractions
  • Natural transitions
  • Some personal voice
  • Occasional idioms

Heavy humanization (casual writing, personal essays):

  • Maximum variation including very short sentences
  • Frequent contractions
  • Personal anecdotes and voice
  • Hedge words and thinking markers
  • Colloquialisms and idioms
  • Self-correction patterns

Important Considerations

Preserve:

  • Original meaning and key arguments
  • Factual accuracy
  • Technical terms when necessary
  • Core structure and organization

Context awareness:

  • Academic writing: lighter touch, keep sophistication, minimal contractions
  • Professional emails: medium touch, maintain clarity and professionalism
  • Blog posts/casual content: heavier touch, maximize personality
  • TEFL assignments: medium-heavy touch, show natural teacher voice

Quality checks:

  • Read aloud - does it sound like someone talking?
  • Check sentence length variation visually
  • Verify no more than 2-3 consecutive sentences start the same way
  • Ensure personality without sacrificing clarity

Reference Files

For detailed guidance:

  • references/ai-tells.md - Comprehensive list of AI overused words/phrases
  • references/linguistic-patterns.md - Research-based differences between AI and human writing
  • references/examples.md - Before/after transformation examples across contexts

Quick Self-Check

After transformation, verify:

  • Sentence lengths vary dramatically (some under 8 words, some over 20)
  • At least 3-5 contractions per 200 words (unless formal context)
  • No "delve," "tapestry," "robust" or similar AI-tells
  • At least one sentence fragment or very short sentence per paragraph
  • Natural transitions (And, But, So) outnumber formal ones
  • Reads naturally when spoken aloud
  • Includes at least one thinking marker or hedge word per 150 words

平台分布

Codex

33.65%
按下载量换算138

Claude

32.83%
按下载量换算135

Cursor

18.74%
按下载量换算77

Gemini CLI

10%
按下载量换算41

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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