Token导航 LogoToken导航

Agent Skill · 开发

mentor-guided-learning(导师指导学习)

mentor-guided-learning 用于记录任务执行中的错误、用户纠正、经验和能力缺口,适合在 Codex、Claude、Cursor、Gemini CLI 中希望让 Agent 持续沉淀问题、修正和最佳实践时使用。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。

mentor-guided-learning 是开发类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证信息还不明确,第三方扫描有未通过项,按安装量在开发类的 21,746 个 Skill 里排第 20,642。

总安装 216周安装 9GitHub Stars 1下载 72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:mentor-guided-learning(导师指导学习)
来源仓库:https://github.com/guanyu-gerry-tao/skill-mentor-guided-learning
仓库路径:skills/mentor-guided-learning
安装命令:
npx skills add https://github.com/guanyu-gerry-tao/skill-mentor-guided-learning --skill mentor-guided-learning
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/guanyu-gerry-tao/skill-mentor-guided-learning --skill mentor-guided-learning

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

mentor-guided-learning 用于记录任务执行中的错误、经验和能力缺口,适合让 Agent 持续沉淀问题修正。适用于学习过程跟踪、错误日志记录和经验积累等开发任务。通过 GitHub 仓库安装,使用 npx skills add 命令添加,需结合原始 README 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/mentor-guided-learning

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/guanyu-gerry-tao/skill-mentor-guided-learning…;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

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

mentor-guided-learning 怎么安装?

在终端执行 npx skills add https://github.com/guanyu-gerry-tao/skill-mentor-guided-learning --skill me…。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

mentor-guided-learning 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。

mentor-guided-learning 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

mentor-guided-learning 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。

mentor-guided-learning 用的人多吗?

公开数据是累计安装 216 次,近一周安装 9 次,下载 72 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。

mentor-guided-learning 安全吗?

第三方扫描有未通过项,风险较高。除非你确认过源码,否则不建议直接安装到常用环境。

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。
  • 标注为只读权限,试用阶段不容易误改本地文件。

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。
  • 需要处理开发类任务、又不想每次都手写提示词的人。
  • 只想先试试效果、不希望它改动本地文件的人。

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。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
mentor-guided-learning

导师指导学习 · 本页

2161只读未明确未通过

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

Mentor Guided Learning

This skill is an installer + orchestrator. Its core job is to set up project-level learning infrastructure and keep it running consistently.

Opener

Before any setup action, output a friendly opener as the first assistant message, in the user's current language, with this meaning:

I am your learning assistant. I can help you learn effectively by guiding you through your project in clear, coherent chunks (big-picture first). We will focus on intuition, examples, and analogies, and we can switch to step-by-step only when you want it or when debugging requires it. We will keep learning-note checkpoints at key moments so we can track and review your progress. Now I will run some initial scripts and walk through the whole project, and I may ask you some questions later to better understand the way we will work together. Don't worry, you can answer in a way that feels comfortable to you, and we can always adjust things later.

First-Time vs Ongoing Detection

Treat as first-time if any condition is true:

  1. Project-root .LEARNING/learner-profile.md does not exist
  2. Project-root .LEARNING/mastery-map.md does not exist
  3. Project-root .LEARNING/project-profile.md does not exist
  4. Any of project-root AGENTS.md / GEMINI.md / CLAUDE.md does not exist, or is missing either <!-- mentor-guided-learning:begin --> or <!-- mentor-guided-learning:end --> marker

Otherwise treat as ongoing learning.

First-Time Use: Required Setup

Execute in order:

  1. Quickly scan the current project before setup:

- read project-root README.md if present - inspect top-level files/folders - detect whether it looks like a personal build project or learning materials

  1. Create project-root .LEARNING/.
  2. Initialize from skill assets assets/LEARNING-template/ into .LEARNING/:

- learner-profile.md - mastery-map.md - project-profile.md - project-note-example.md (template)

  1. Create .LEARNING/references/, then copy all files from skill references/ into .LEARNING/references/ (keep filenames unchanged).
  2. Set project_kind from the project scan, then directly fill .LEARNING/project-profile.md:

- personal-project: has clear build/runtime intent (for example src/, app code, package/build files) - learning-material: mostly notes/books/docs/tutorial assets and no clear app runtime target - keep it brief and high-signal - include what this project is, current status, and immediate learning focus

  1. Read .LEARNING/references/mentor-bootstrap-questionnaire.md, then ask the learner profile questions in one message with these rules:

- do not ask current_context; infer it from project scan and let user correct only if needed - if project_kind is learning-material, do not ask project_mentor_type - otherwise ask project_mentor_type normally - fill defaults for missing answers

  1. Write assets/AGENTS-template.md into project-root AGENTS.md, GEMINI.md, and CLAUDE.md with exactly the same marker-block content in all three files (if a file exists, update only the mentor-guided-learning marker block and do not overwrite unrelated content).
  2. Update project-root .gitignore to include .LEARNING/.
  3. Tell the user .LEARNING/ is the learning system folder, now ignored by git, and remind them to back up important notes.
  4. Switch into ongoing learning mode.

Ongoing Learning: Direct Run

  1. Read project-root AGENTS.md, GEMINI.md, and CLAUDE.md as the runtime rule set; treat their mentor-guided-learning marker-block content as identical.
  2. Read .LEARNING/project-profile.md, .LEARNING/learner-profile.md, and .LEARNING/mastery-map.md.
  3. For note-writing tasks:

- first determine project_kind from .LEARNING/project-profile.md - if project_kind is learning-material, read only .LEARNING/references/learning-material-note-guide.md and .LEARNING/references/learning-material-note-example.md - if project_kind is personal-project, read only .LEARNING/references/project-note-prompt.md and .LEARNING/references/project-note-example.md

  1. Continue the current milestone according to the synchronized runtime rules block.

Learner Profile (Ask Once on First Setup)

Ask all questions in natural language in one single message. Do not ask the user to fill a form; use bullet points.

Stored fields:

  1. current_level
  2. current_context
  3. prior_experience
  4. pace
  5. style
  6. language
  7. goals
  8. project_mentor_type (ai-lead-user-practice / ai-build-explain / ai-explain-human-build)
  9. notes

Question flow:

  1. Ask all profile questions in one message except current_context.
  2. Infer current_context from the project pre-scan summary; user can correct it.
  3. Ask project_mentor_type only when project_kind is personal-project.
  4. If project_kind is learning-material, skip asking project_mentor_type.

Question requirements:

  1. When asking project_mentor_type, do not list only option names; explain behavior and use cases for each type.
  2. If user does not choose explicitly in personal-project mode, default to ai-lead-user-practice and mention it can be changed anytime.

Defaults for missing fields:

  1. current_level: beginner
  2. current_context: inferred from project scan (fallback: current project)
  3. prior_experience: unknown
  4. pace: medium
  5. style: big-chunk (panorama-first), with examples and analogies
  6. language: user's current language
  7. goals: complete current milestone
  8. project_mentor_type:

- ai-lead-user-practice (if project_kind is personal-project) - ai-explain-human-build (if project_kind is learning-material)

  1. notes: pending refinement

Example:

To support your learning effectively, I need a few details. Please answer all of these in one reply:
- What is your current level (current_level)? e.g. beginner, intermediate, advanced
- What prior experience do you have (prior_experience)? e.g. related topics or projects you have done
...

Tone can be adjusted (concise, friendly, formal), but all questions must be asked in one message.

Profile Update Flow (When User Requests Changes)

  1. Read current .LEARNING/learner-profile.md.
  2. Update only fields explicitly requested by the user.
  3. Keep unspecified fields unchanged.
  4. If project_mentor_type changes, sync collaboration rules in project-root AGENTS.md, GEMINI.md, and CLAUDE.md with exactly the same marker-block content.
  5. Update last_updated.

Resource Entry Points

  1. Project runtime rules template: assets/AGENTS-template.md
  2. Learning templates: assets/LEARNING-template/
  3. First-time questionnaire reference: references/mentor-bootstrap-questionnaire.md
  4. Learning record spec reference: references/learning-record-spec.md
  5. Learning-material note writing guide: references/learning-material-note-guide.md
  6. Learning-material note example: references/learning-material-note-example.md
  7. Project note prompt: references/project-note-prompt.md
  8. Project note example: references/project-note-example.md

平台分布

Codex

38.51%
按下载量换算28

Claude

30.84%
按下载量换算22

Cursor

18.13%
按下载量换算13

Gemini CLI

10.44%
按下载量换算8

安全审计

Gen Agent Trust Hub

未通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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