Token导航 LogoToken导航

Agent Skill · 开发规范

human-guidelines(人类指导方针)

human-guidelines 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。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-guidelines 是开发规范类 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,第三方扫描有提醒项,按安装量在开发规范类的 1,265 个 Skill 里排第 509。

总安装 1,211周安装 52GitHub Stars 公开资料未说明下载 424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-guidelines(人类指导方针)
来源仓库:https://github.com/heyimjames/human-guidelines
仓库路径:skills/human-guidelines
安装命令:
npx skills add https://github.com/heyimjames/human-guidelines --skill human-guidelines
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/heyimjames/human-guidelines --skill human-guidelines

这个 Skill 是做什么的

开发规范类 Skill 用来统一代码风格、目录结构和评审口径,产出多为检查清单或改写建议。

human-guidelines 用于处理 GitHub 仓库、Issue 和 Pull Request 信息,适合项目状态整理。适用于开发规范制定和协作事项管理场景。通过 npx skills add 命令从指定仓库安装,需确认权限范围。使用前建议核实维护状态及是否涉及敏感数据操作。

可结合原始 SKILL.md 了解具体交互方式和输出格式。

来源仓库内路径:skills/human-guidelines

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/heyimjames/human-guidelines --skill human-guid…;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-guidelines 怎么安装?

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

human-guidelines 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-guidelines 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-guidelines 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-guidelines 用的人多吗?

公开数据是累计安装 1,211 次,近一周安装 52 次,下载 424 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。

human-guidelines 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

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。

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。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 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。
  • 需要处理开发规范类任务、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。

开发规范类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
vercel-react-best-practices

Vercel React 最佳实践 · Codex、Claude

8,490,72025,918操作浏览器MIT通过
web-design-guidelines

网页设计指南 · Codex、Claude

6,686,01625,885需要联网MIT有提醒
remotion-best-practices

Remotion 最佳实践 · Codex、Claude

6,505,6323,027操作浏览器MIT有提醒
supabase-postgres-best-practices

Supabase Postgres 最佳实践 · Codex、Claude

3,175,5232,030需要联网MIT通过
next-best-practices

Next.js 最佳实践 · Codex、Claude

1,848,240841操作浏览器MIT通过
human-guidelines

人类指导方针 · 本页

1,211—需要联网未明确有提醒

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 Guidelines

Make AI-generated content feel like a human created it. Covers UI design, writing, and social media.

Guidelines Source

Fetch fresh guidelines from guidelines.sh before each task:

  • UI: https://guidelines.sh/api/guidelines?type=ui
  • Writing: https://guidelines.sh/api/guidelines?type=writing
  • Social: https://guidelines.sh/api/guidelines?type=social
  • All: https://guidelines.sh/llms-full.txt

Use WebFetch to retrieve the latest rules.


UI Guidelines

Make interfaces feel crafted by humans, not generated by algorithms.

Interview Mode

When the user says "interview", "help me define my taste", or "what's my style":

Run through these questions one at a time. Wait for answers. Build a taste profile.

  1. Pick 3 adjectives for how your product should feel
  2. What apps do you love using? (for how they feel, not features)
  3. What do you hate in interfaces?
  4. Spatial preference: Dense or spacious?
  5. Animation personality: Snappy or playful?
  6. Color mood: Monochromatic? Warm? Cool? Bold?
  7. Typography character: Neutral or distinctive?
  8. What's the one thing that would make it feel "you"?

Output a Taste Profile summary they can save.

UI Quick Reference

Taste & Craft

  • Taste is having opinions—AI has none
  • Collect references obsessively
  • Define your vibe and commit
  • Handcrafted details compound
  • Prioritize feel over features (in product UI, "feel" = responsiveness and flow; on marketing pages, "feel" = mood and first impressions)

Micro-interactions

  • Animation is personality—choose deliberately
  • Use spring physics over bezier curves
  • Stagger list reveals (30-50ms delay)
  • Design exits as carefully as entrances
  • Every interactive element needs hover feedback
  • Scale animation to context: marketing can be dramatic; product UI should be invisible by the 5th view

Key Checks

  • Dark mode: Never pure black (#000000), use #1a1a1a
  • Typography: text-wrap: pretty, vary fonts
  • Animation: Organic easing, imperfect timing
  • Layout: Asymmetry for marketing; consistent grids for product UI
  • Icons: Three states (stroke/duotone/solid)
  • Density: Marketing wants space; product UI depends on use case (dense for dashboards, spacious for onboarding)

Surface-Specific Guidance

Marketing & Landing Pages:

  • First impression is everything—hero, scroll pacing, narrative flow
  • Brand expression lives here—be brave with creative risks
  • Social proof needs craft—real quotes, specific numbers, genuine photography
  • CTAs should feel inevitable, not pushy
  • Performance is invisible design—Core Web Vitals affect SEO and trust

In-App Product UI:

  • Consistency beats novelty—users build muscle memory
  • Design for the 100th session, not the 1st
  • States are the real design work—loading, empty, error, partial, offline
  • Power users are watching—keyboard shortcuts, bulk actions, density toggles
  • Data is content—tables, charts, and lists deserve the same care as a hero section
  • Template patterns are a feature here—the craft is in execution quality, not structural novelty
  • Empty states are your biggest opportunity—they're the first thing new users see

Writing Guidelines

Make AI-generated writing feel like a human wrote it.

Writing Quick Reference

Priority order for humanizing text:

  1. Kill preamble - Remove "Certainly!", "Great question!", "I'd be happy to help"
  2. Ban AI vocabulary - Replace "delve", "utilize", "leverage", "facilitate", "comprehensive"
  3. Use contractions - "don't" not "do not", "it's" not "it is"
  4. Vary sentence rhythm - Mix short punchy sentences with longer flowing ones
  5. Get specific - Replace vague claims with concrete details
  6. Cut filler - Remove "It's important to note that", "In order to"
  7. Sound opinionated - Take stances instead of hedging everything

Writing Usage

  1. Analyze text against all rules
  2. List specific violations found
  3. Provide rewritten version with fixes applied

Social Media Guidelines

Make social posts feel like a person sharing, not a brand broadcasting.

Social Quick Reference

Universal Fixes

  1. Remove hashtag walls - Keep 0-2 max, or none
  2. Delete corporate openers - "I'm thrilled to announce", "Excited to share"
  3. Cut emoji bullets - Those checkmark/rocket/star lists are AI tells
  4. Use lowercase - Sentence case feels more human
  5. Strategic imperfection - Occasional typo is okay
  6. Be specific - "shipped the thing" > "successfully launched our innovative solution"

Platform-Specific

X/Twitter: Hot takes, no hashtag walls, embrace brevity LinkedIn: Specific wins, no "Agree?" prompts, real stories Instagram: Visual-first, short captions TikTok: Authentic moments, hook in 2 seconds

Social Usage

  1. Ask which platform if not specified
  2. Apply platform-specific rules
  3. Output post with humanization applied
  4. Offer 2-3 variations if helpful

平台分布

Codex

32.58%
按下载量换算138

Claude

28.44%
按下载量换算121

Cursor

19.99%
按下载量换算85

Gemini CLI

10.19%
按下载量换算43

安全审计

Gen Agent Trust Hub

可疑

Socket

通过

Snyk

可疑

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。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