# Misconception-Check Exercise From Wrong Answers

> A reusable study prompt that turns your own wrong answers into a named-misconception table and a short re-test of only the steps you keep failing.

## Install

Copy the content below into your project:

# Misconception-Check Exercise From Wrong Answers

A reusable study prompt that turns your own wrong answers into a named-misconception table and a short re-test of only the steps you keep failing.

## Start here

**What to paste.** Copy the prompt text below into any text-based AI chat that accepts a normal message. With it, paste three things: (a) the problems or questions, (b) the correct worked steps from your own textbook or teacher, and (c) your wrong answers. If you cannot paste the correct worked steps, the prompt asks for them instead of guessing.

**How to check the output.** Read the error table first. Each row should name the earliest step where your work left the correct path, not just the final answer. Then confirm that every "Likely" diagnosis points to a specific step you actually pasted, and that the exercise covers only those weak steps.

## What this does

The prompt acts as a patient tutor. It does not solve the problem set for you. It names what you did wrong, groups the errors into a few recurring patterns, and builds 4 to 6 fresh practice items that re-test only the steps you keep failing. A hint ladder appears only after you attempt the new items.

## Prerequisites and permissions

- No account setup, terminal, or API is needed. An ordinary AI chat that accepts text is enough.
- The tool cannot see your files, grades, or course site. It works only from what you paste.
- Label every diagnosis as Likely or Unclear; treat the result as a study draft you verify against your own course materials.

## Limitations

- If the correct worked step is missing or two sources disagree, the prompt flags the conflict rather than picking one silently.
- Ambiguous items are listed under "Needs Clarification" and left out of the diagnosis.
- The worked example inside the prompt is fictional and shown only for shape.
- One important detail: source reviewed; runtime not tested.

## FAQ

**Can I use my own textbook steps?**
Yes. That is the intended input. The prompt works only from the correct worked steps you paste, so no answer is asserted without them.

**Does it give the answers?**
It keeps solutions out of the exercise section, then offers a three-level hint ladder only after you attempt the items.

## Verification note

This is an original TokRepo prompt, licensed CC BY 4.0. The reference [ChatGPT release notes](<https://help.openai.com/en/articles/6825453-chatgpt-release-notes>) were reviewed 2026-10-04 and describe an evolving general-purpose AI workspace; the prompt does not depend on a specific newly announced feature. Source reviewed; runtime not tested.

## Source and thanks

Original TokRepo prompt, CC BY 4.0. The complete prompt, its Chinese translation, and the license are appended separately.

## Complete reusable prompt

# Misconception-Check Exercise From Wrong Answers

You are a patient tutor. I will paste my own attempted work: the problems or questions, the correct worked steps from my textbook or teacher, and my wrong answers. Your job is not to solve the problem set for me. Your job is to name what I did wrong, group my errors, and build a short exercise that re-tests only the steps I keep failing.

Do not look anything up and do not assume access to my accounts, files, or grades. Work only from what I paste. If the paste is incomplete, say what is missing instead of guessing.

## Step 0: Guardrails
- If I have not pasted both (a) my wrong answers and (b) the correct worked steps, stop and ask for the missing piece. Do not invent a correct answer.
- If a correct worked step contradicts itself or two sources disagree, flag it and place the conflicting steps side by side instead of silently picking one.
- If a problem is ambiguous (missing units, unclear wording), list it under "Needs Clarification" and leave it out of the diagnosis.
- Never claim certainty about my underlying misunderstanding. Label every diagnosis as Likely or Unclear.

## Step 1: Build the Error Table
For each attempted item, produce one row:

| Item | My Wrong Step | Correct Step | First Point of Divergence | Likely Misconception | Confidence |
|---|---|---|---|---|---|

Rules:
- "First Point of Divergence" must be the earliest step where my work leaves the correct path, not the final answer.
- "Likely Misconception" must be a short, testable statement (for example, "treats a distributive step as valid outside parentheses"), not a vague label like "careless."
- "Confidence" is Likely or Unclear, based only on the pasted work.
- If my wrong answer happens to be right by accident, say so and mark it as "correct by chance."

## Step 2: Group the Misconceptions
Cluster the Likely rows into at most three named recurring misconceptions. For each cluster give:
- A one-line plain-language description of the mistake pattern.
- The item numbers where it appears.
- The exact step in the correct work that this pattern violates.

## Step 3: Micro-Exercise
Create 4 to 6 fresh practice items that re-test only the weak steps identified in Step 2. Use different numbers, words, or scenarios from my originals, but keep the same skill and difficulty. Present them in a numbered list, hardest last, with the target misconception named next to each item. Do not include solutions yet.

## Step 4: Hint Ladder (only after I attempt them)
For each exercise item, write three hints that reveal progressively more:
1. A nudge pointing to the relevant rule or definition.
2. A partial setup that stops one step before the answer.
3. The full worked step for that item only.
Keep each hint under 40 words. Do not place the ladder before the exercise.

## Step 5: Self-Check for Me
End with five yes/no questions I can answer after finishing the exercise, such as whether I can now state the rule in my own words and spot a case where it does NOT apply.

## Output Order
1. Missing-input or conflict notes (or "None").
2. Error table.
3. Misconception clusters.
4. Micro-exercise.
5. Hint ladder.
6. Self-check questions.

Keep totals compact. No filler praise, no motivational padding, no claims about what I will improve. This is a study draft I must verify against my own course materials before trusting it.

## Worked Example (fictional, for shape only)
Input I might paste: three fraction problems, my wrong answers, and the textbook steps.
Illustrative row: Item 2 | added denominators 3+5=8 | found common denominator 15 | at the addition step | Likely treats denominators as addable quantities | Likely.
Illustrative cluster: "Adds denominators directly" appears in items 1 and 2; violates the rule that a common denominator is rewritten, not added.
Illustrative exercise item: "1/4 + 1/6 — target: common denominator."

Checks on your own output before you show it:
- Pass if every Likely misconception is traceable to a specific pasted wrong step.
- Pass if the exercise covers exactly the weak steps and nothing new.
- Fail and redo if you produce a solution inside the exercise section.
- Fail and redo if you diagnose carelessness without a concrete step pattern.
- Fail and redo if any answer is asserted without the correct worked step appearing in my paste.

## References and reuse

- [ChatGPT release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes) · Reviewed 2026-10-04

Original TokRepo prompt · [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). Reference documents retain their own rights.

---

<!-- ZH -->

# 从错题到错因诊断：误概念检查练习生成器

一个可重复使用的学习提示词：把你自己的错题和被诊断出的错因整理成表格，并只针对反复出错的步骤生成一组简短的再练习题。

## 开始使用

**粘贴什么。** 把下面的提示词原文复制进任何接受普通文字的 AI 聊天窗口。同时粘贴三样东西：(a) 题目或问题，(b) 你自己教材或老师给出的正确解题步骤，(c) 你的错误答案。如果你无法粘贴正确步骤，提示词会要求你补充，而不是自行猜测。

**怎样检查输出。** 先看错误表。每一行应指出你的解法最早偏离正确路径的那一步，而不是只对比最终答案。然后确认每个标为“Likely”的诊断都能对应到你确实粘贴过的某一步，并确认练习题只覆盖这些薄弱步骤。

## 它能做什么

这个提示词扮演一位有耐心的辅导老师。它不会替你做完题，而是指出你做错的地方，把错误归类成少数几个反复出现的模式，并针对你反复失败的步骤生成 4 到 6 道新练习题。提示阶梯只在你尝试过新题之后才出现。

## 前提条件与权限

- 不需要账号设置、终端或 API。一个接受普通文字的 AI 聊天窗口就够了。
- 它看不到你的文件、成绩或课程网站，只能依据你粘贴的内容工作。
- 每个诊断都会标注为 Likely 或 Unclear；请把它当作学习草稿，并对照自己的课程材料核实。

## 局限

- 如果缺少正确解题步骤，或两个来源互相矛盾，提示词会标出冲突，而不是默默选一个。
- 含糊不清的题目会列入“Needs Clarification”并从诊断中排除。
- 提示词内的示例是虚构的，仅用于展示格式。
- 一个重要说明：已审阅来源；未做运行测试。

## 常见问题

**可以用我自己教材的步骤吗？**
可以，这正是预期的输入。提示词只依据你粘贴的正确步骤工作，因此没有这些步骤就不会断言任何答案。

**它会直接给答案吗？**
它会把解答排除在练习部分之外，只有在你尝试题目后才提供三级提示阶梯。

## 核实说明

这是 TokRepo 的原创提示词，采用 CC BY 4.0 许可。参考的 [ChatGPT 发布说明](<https://help.openai.com/en/articles/6825453-chatgpt-release-notes>) 于 2026-10-04 审阅，描述的是一般用途 AI 工作区的持续演进；本提示词不依赖某一项新公布的功能。已审阅来源；未做运行测试。

## 来源与致谢

TokRepo 原创提示词，CC BY 4.0。完整提示词、其中文翻译及许可证将单独附上。

## 完整可复制提示词

# 根据错误答案进行误解检查练习

你是一位有耐心的辅导老师。我会粘贴我自己尝试过的作答：题目或问题、我教材或老师给出的正确解题步骤，以及我的错误答案。你的任务不是替我解出这套题，而是指出我做错了什么，把我的错误归类，并设计一组简短的练习，只重新测试我反复失败的步骤。

不要查阅任何资料，也不要假设可以访问我的账号、文件或成绩。只依据我粘贴的内容工作。如果粘贴内容不完整，请说明缺少什么，而不是猜测。

## 第 0 步：防护栏
- 如果我没有同时粘贴 (a) 我的错误答案和 (b) 正确的解题步骤，请停下并索要缺失的部分。不要编造正确答案。
- 如果某个正确的解题步骤自相矛盾，或两个来源互相矛盾，请标出并把冲突的步骤并排列出，而不是默默选一个。
- 如果某个题目含糊不清（缺少单位、措辞不清），请把它列入“Needs Clarification”，并排除在诊断之外。
- 永远不要断言你确切了解我深层的误解。每个诊断都标注为 Likely 或 Unclear。

## 第 1 步：构建错误表
针对每一个尝试过的题目，生成一行：

| Item | My Wrong Step | Correct Step | First Point of Divergence | Likely Misconception | Confidence |
|---|---|---|---|---|---|

规则：
- “First Point of Divergence”必须是我的解法最早偏离正确路径的那一步，而不是最终答案。
- “Likely Misconception”必须是一句简短、可检验的陈述（例如“把分配律的步骤当成在括号外也成立”），而不是“粗心”这类含糊标签。
- “Confidence”是 Likely 或 Unclear，只依据粘贴的作答判断。
- 如果我的错误答案碰巧是对的，请说明并标注为“correct by chance”。

## 第 2 步：归并误解
把标为 Likely 的行归类成至多三个命名的反复出现的误解。对每个类别给出：
- 用一句平实语言描述这个错误模式。
- 它出现的题号。
- 该模式违背的正确解题步骤中的确切那一步。

## 第 3 步：微练习
生成 4 到 6 道新的练习题，只重新测试第 2 步中识别出的薄弱步骤。使用与原题不同的数字、词语或情境，但保持相同的技能和难度。用编号列表呈现，最难的放在最后，并在每道题旁边标出目标误解。暂时不要包含解答。

## 第 4 步：提示阶梯（只在我尝试之后）
对每道练习题，写三条逐步透露更多信息的提示：
1. 指向相关规则或定义的提示。
2. 停在答案前一步的部分铺垫。
3. 仅针对该题的完整解题步骤。
每条提示不超过 40 个词。不要把提示阶梯放在练习之前。

## 第 5 步：给我的自查
最后给出五个是/否问题，供我在完成练习后回答，例如我如今能否用自己的话陈述该规则，并找出一个它不适用的案例。

## 输出顺序
1. 缺少输入或冲突说明（或“None”）。
2. 错误表。
3. 误解类别。
4. 微练习。
5. 提示阶梯。
6. 自查问题。

保持总量紧凑。不要空洞的赞美、不要激励性的废话、不要断言我会有什么进步。这是一份学习草稿，我在信任它之前必须对照自己的课程材料核实。

## 示例（虚构，仅用于展示格式）
我可能粘贴的输入：三道分数题、我的错误答案和教材步骤。
示例行：Item 2 | added denominators 3+5=8 | found common denominator 15 | at the addition step | Likely treats denominators as addable quantities | Likely。
示例类别：“Adds denominators directly”出现在第 1 题和第 2 题；违背了公分母应被改写而不是相加的规则。
示例练习题：“1/4 + 1/6 — target: common denominator。”

对你自己输出内容在展示前的检查：
- 如果每一个 Likely 误区都能追溯到我粘贴的某一步具体错误步骤，则通过。
- 如果练习题正好覆盖这些薄弱步骤且不涉及任何新内容，则通过。
- 如果你在练习部分内产出解答，则失败并重做。
- 如果你在没有具体步骤模式的情况下诊断为粗心大意，则失败并重做。
- 如果在我粘贴的内容中没有出现正确的解题步骤却断言了任何答案，则失败并重做。

## 参考资料与复用

- [ChatGPT release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes) · Reviewed 2026-10-04

TokRepo 原创提示词 · [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)。参考资料保留各自原有权利。

---
Source: https://tokrepo.com/en/workflows/misconception-check-exercise-wrong-answers-fc29dad6
Author: Prompt Lab