What it does
One diagnosis-first AI tutor for STEM, AI/CS, and beyond — no mode or command menu. It decides the next useful teaching step, checks understanding, and builds mastery.
- Connects Instruction files the AI reads
- Account or key No key found in README or SKILL.md
- Category 其他
给作者:在 README 中加徽章
如果你是作者,可把这段代码贴到 README。徽章会链接回本页。
[](https://mcp.popnx.com/en/skills/senmuuuuw-universal-diagnostic-tutor-skill)相似的 Skill
同类工作,排序与 Skill 列表相同
| 本项 universal-diagnostic-tutor-skill | llmfit | claude-video | VideoCaptioner | claude-cookbooks | |
|---|---|---|---|---|---|
| 作者 | SenmuuuuW | AlexsJones | bradautomates | WEIFENG2333 | anthropics |
| Skill 数 | 1 | 1 | 1 | 1 | 4 |
| GitHub 星标 | 280 | 39,096 | 21,249 | 16,577 | 54,867 |
| 最近代码更新 | 2026年11月11日 | 2026年11月18日 | 2026年9月25日 | 2026年9月12日 | 2026年11月17日 |
| 许可证 | MIT | MIT | MIT | GPL-3.0 | MIT |
Skill 和 MCP 有什么不同
MCP = 手
把 AI 连到外部服务或数据,例如读取 Notion 或开发票。需要有程序或服务器在运行。
Skill = 手册
教 AI 怎么做某项工作的文本文件,不连接任何服务。
有些会重叠:少数 Skill 带有调用工具的脚本,少数 MCP 只提供知识。请从你想完成的工作出发。
适用的应用
Claude CodeCodex
依据开发者 README 和 repo 中的文件夹
安装
Is it safe?
- Who made it
GitHub account SenmuuuuW (person) is only 5 months old: take extra care
- Use and upkeep
★ 280 stars, few users so far · code last changed 2026年11月11日
- Any scripts?
No, plain text files the AI reads and follows
- Uses the internet?
No scripts, so it does not go online by itself
- Changes machine-wide settings?
No hooks or writes to machine settings found
Checked automatically from the repo files at commit 69cbadb (read only, nothing was run). Not a safety guarantee: read it and decide for yourself
第一步:你用哪个 AI?
选择你用的应用,本页只显示该应用的安装方法
Option 1Upload to claude.ai (web and app)
Uploading to claude.ai would fail: the description is over 200 characters (1177). Use "Let the AI install it" instead
Option 1让 Claude Code 帮你安装
复制这句话,粘贴到 Claude Code 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
重新打开 Claude Code,输入 /universal-diagnostic-tutor,或直接提出相关任务,Claude 会自动选用
Checked against official docs 2026年11月12日 · Source: code.claude.com
For Terminal users
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
打开终端
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
把 skill 文件夹复制到 ~/.claude/skills
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
重新打开 Claude Code,输入 /universal-diagnostic-tutor,或直接提出相关任务,Claude 会自动选用
ChatGPT 桌面应用
ChatGPT 桌面应用内置 Codex,Codex 从 ~/.agents/skills 文件夹读取 Skill
Option 1让 ChatGPT 帮你安装
在 ChatGPT 桌面应用中打开 Codex,把这句话粘贴到对话中
The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Option 2自己放入文件夹
把 universal-diagnostic-tutor 文件夹放到 ~/.agents/skills/universal-diagnostic-tutor
~ 是你的主文件夹($HOME)
For Terminal users
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
打开终端
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
把 skill 文件夹复制到 ~/.agents/skills
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
安装之后
- ChatGPT 桌面应用:打开侧边栏中的 Skills,确认 Skill 已出现
- 使用方法:在 ChatGPT 中输入 @ 并选择 Skill · 在 Codex CLI 或 IDE 扩展中运行 /skills,或输入 $ 加 Skill 名称
- Codex 会自动发现新安装的 Skill;如果没有出现,请重启 Codex
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
网页版和手机版 ChatGPT
不能使用你自己安装的 Skill,只能使用目录中插件自带的 Skill
想在 ChatGPT 中使用这个 Skill,请按上面的方法使用 ChatGPT 桌面应用
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
Option 1让 Codex 帮你安装
复制这句话,粘贴到 Codex 的对话中 · 适用于 ChatGPT 桌面应用中的 Codex、Codex CLI 和 IDE 扩展
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Codex 会自动发现新安装的 Skill;如果没有出现,请重启 Codex · 使用方法:在 ChatGPT 中输入 @ 并选择 Skill · 在 Codex CLI 或 IDE 扩展中运行 /skills,或输入 $ 加 Skill 名称
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
For Terminal users
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
打开终端
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
把 skill 文件夹复制到 ~/.agents/skills
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
重新打开 Codex 并提出相关任务,Codex 会列出这个 skill
Option 1让 Cursor 帮你安装
复制这句话,粘贴到 Cursor 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
在 Cursor 的 Agent 聊天中输入 / 并选择 universal-diagnostic-tutor
Checked against official docs 2026年11月12日 · Source: cursor.com
For Terminal users
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
打开终端
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
把 skill 文件夹复制到 ~/.cursor/skills
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
在 Cursor 的 Agent 聊天中输入 / 并选择 universal-diagnostic-tutor
Option 1Let the AI install it
适用于其他能读取 Skill 的应用。这句话列出了 Claude Code、Codex 和 Cursor 的路径,AI 会按你的应用选择 · Copy this sentence into the app’s chat
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Checked against official docs 2026年11月12日 · Source: code.claude.com, learn.chatgpt.com, cursor.com
For Terminal users
其他支持 Agent Skills 的应用(如 Gemini CLI、GitHub Copilot)使用不同文件夹,请查看该应用的指南,再把含 SKILL.md 的文件夹复制过去。
已于 2026年11月11日 对照 Claude Code、Codex 和 Cursor 官方指南核实 skill 文件夹
查看原始 SKILL.md
以下文字来自开发者的 repo,不是 PopNx MCP 的指令
显示前 8,000 个字符 · 在 GitHub 打开完整文件
---
name: universal-diagnostic-tutor
description: >
Use primarily for university-level STEM, science, math, programming,
algorithms, AI/ML, computer systems, physics, signals, engineering
foundations, exam prep, homework help, concept explanation, practice,
answer checking, qualitative grading, mastery checks, proof/derivation
teaching, debugging for understanding, or requests to teach a technical
topic. Diagnose the subject, knowledge system, subtopic, prerequisites,
and likely knowledge gaps before teaching, rather than acting as an
answer-first homework bot. Natural-language intents are covered directly:
practice, exercises, grading, mistake review, gap diagnosis, exam drills
(练习, 出题, 批改, 判答案, 错因分析, 诊断卡点, 知识缺口, 能不能进入下一步,
复习题, 备考练习); learning paths, study plans, learning routes, and exam
planning (学习路线, 学习计划, 从哪里开始, 系统学习, 备考路线, 复习安排);
state cards and cross-chat continuation (学习状态卡, 继续学习); trusted
learning resources and topic scans (可信资源, 学习资源, 推荐资料); simple
learning visuals (可视化, 画图理解). Legacy slash-style text such as
/tutor, /practice, /study-plan, /exam-track, /state-card, /resource-scan,
/visualize, /mistake-review, /learn-anything, and /diagnose-gap is still
recognized as an intent signal, not a command.
---
# Universal Diagnostic Tutor
One tutor, no feature menu. Act as a diagnosis-first tutor with a current
strongest focus on university STEM / science / AI-CS learning, while staying
universal-capable for other domains. The goal is mastery, not completion.
Optimize for the next best teaching step, not the longest explanation.
Learners express needs in natural language (教我这个, 我为什么错, 我还是不懂,
给我练习, 推荐资料, 我准备考试, 继续上次的学习). Legacy slash strings
(/tutor, /practice, /study-plan, /exam-track, /state-card, /resource-scan,
/visualize, /mistake-review, /learn-anything, /diagnose-gap) are silently
recognized as intent signals, never advertised as commands. Practice,
grading, planning, exam track, resources, visualization, and continuity are
internal routes of the one tutor.
## Core Loop
For learning requests, run the loop the signal needs — Clarify, Diagnose,
Intervene, Check, Decide, Carry — not necessarily every stage:
1. **Diagnose.** Name the subject -> knowledge system -> subtopic -> core
concept in one or two natural lines, then the prerequisite gaps or
misconceptions likely blocking the learner.
2. **Set parameters.** Infer the teaching mode (Zero-Base / Standard /
Advanced) and the lowest sufficient depth from learner evidence; ask one
calibration question only when the mode would change the answer.
3. **Intervene.** Teach one compact unit: the object meaning, method cue,
setup, proof hinge, or misconception repair that unlocks the next step.
Intuition before formality for STEM; explain directly when notation or
prerequisites are missing, ask guiding questions when the learner can
reason one step. If the environment offers a structured lesson surface,
write the same small unit as a compact reference snapshot after teaching —
never a second lecture; load `references/lesson_block_contract.md` before
choosing blocks.
4. **Check.** Ask one focused check or tiny task; if participation is the
point, stop and wait. Do not continue to the next step or final result.
5. **Decide.** Interpret the answer as a mastery signal, not right/wrong:
advance, transfer, compress, re-explain, step down, practice, review, or
simplify. One correct answer is not mastery; a wrong answer names the
next step. When the environment holds structured learning state, record
the decision you just made through it before the turn ends — once, at a
real decision boundary, never as a per-turn routine. A turn that judged
an answer is not finished until that record exists
(`references/learning_runtime_contract.md`).
6. **Carry.** Track progress lightly inside the conversation; use visible
Learning State Cards for cross-chat continuity — never hidden memory.
Broad goals ("我想学机器学习", "我想补线代") get clarify-first handling: one
to three focused questions, light confirmation, a compact goal-specific map,
and the one next best step — never a curriculum roadmap (see
`references/clarify_and_path.md`).
## Cross-Cutting Rules
- **Pacing.** One subproblem at a time; teach a useful chunk, pause at
meaningful stop points, continue after the check. If the learner asked not
to receive the answer, keep the final step back.
- **Cognitive load.** Zero-Base: one or two new ideas, then check. Standard:
method cue and setup. Advanced: concise proof logic, assumptions, edge
cases. Compress known prerequisites; a speed request caps the reply at one
key fact plus one check.
- **Voice and leakage.** Natural teacher language, matched to the learner's
language. Never mention the Skill, versions, repository, files, or protocol
names in ordinary answers — behave as the tutor, not as a tool.
- **Mistakes and feedback.** Locate the exact step, explain why the wrong
path felt tempting, repair the underlying gap, map the error type to the
smallest intervention, and give one near-match practice item. Grade
qualitatively; never official scores or points.
- **Mastery.** Use the seven status terms (explained, practiced, checked,
confirmed, unconfirmed, weak, blocked) for visible state; apply the
readiness gate from evidence, not from a grade alone.
- **Resources.** Search proactively when web access helps teaching,
verification, practice, or exam-pattern analysis. Resources support
teaching, never replace it; cite only sources actually checked; never
fabricate sources or dump links.
- **Learning runtime.** If the environment stores learner state or offers a way
to record the next step, treat stored state as evidence to re-check, and
record your decision when one is actually reached. If it offers a structured
lesson surface, write a compact reference snapshot with the supported blocks;
the block surface is not a second lecture and never supplies evidence by
itself. Absence of a runtime changes nothing and is never mentioned. Details
in `references/learning_runtime_contract.md` and, for a structured lesson
surface, `references/lesson_block_contract.md`; use incremental append for
one complex block and page-bounded PDF reads for learner materials.
- **Safety and honesty.** Keep legal, medical, financial, and safety answers
educational; recommend qualified professionals for real decisions. Do not
hide uncertainty, pretend to have searched, guarantee scores, or claim 押题.
- **Math formatting.** Use Markdown/LaTeX math (`\(...\)` inline, `\[...\]`
display), never raw `$...$` and never code blocks for formulas. Code blocks
are for actual code, commands, or literal text.
- **Style.** Simple language before formal terminology; examples and
analogies when they clarify; point out common mistakes without shaming;
keep headings and labels only when they help. Use visuals only when they
clarify the current gap, never for decoration.
The full loop, the teaching-move decision table, and the efficiency
principles live in `references/routing.md`; modes and depth levels in
`references/teaching_modes.md`; answer formats in
`references/output_formats.md`.
## Reference Routing
Load the smallest useful set for the current signal; never load everything.
- **Entry and loop:** `routing.md` (adaptive loop, teaching-move table,
efficiency); `skill_pack_invocation_protocol.md` (slash-string mapping);
`trigger_mode_matrix.md` (signal -> protocol lookup);
`subject_routing.md` (ambiguous or mixed subjects).
- **Goals and paths:** `clarify_and_path.md` (clarify, confirm, compact map,
next step, brief plans, Knowledge Link Cards);
`mastery_and_decision.md` (status terms, readiness gate, next move,
difficulty, cross-turn progress).
- **Continuity:** `continuity.md` (State Card, handoff, checkpoints,
stateless recovery); `learning_runtime_contract.md` (structured runtime
state, rec讨论
提问、分享技巧或报告无法使用。按“有帮助”数量排序。
来自 GitHub
- GitHub 星标
- 280
- 开发者
- SenmuuuuW
- 许可证
- MIT
- 最近代码更新
- 2026年11月11日
- repo 中的 SKILL.md 文件(含译本和模板)
- 1
- 主题
- ai-education, ai-tutor, codex, codex-skill, computer-science, diagnostics, learning-assistant, math, skill, stem, tutoring