# Prompt: Turn Rough Notes into a Clear How-To Guide

> An original TokRepo prompt that turns your rough step notes into a numbered how-to guide, flagging every gap and asking you to fill it.

## Install

Copy the content below into your project:

# Prompt: Turn Rough Notes into a Clear How-To Guide

An original TokRepo prompt that turns your rough step notes into a numbered how-to guide, flagging every gap and asking you to fill it.

## Start here

**What to paste:** one message containing your GUIDE GOAL (one sentence), READER PROFILE (who follows the steps, their skill level, assumed tools), RAW STEP NOTES (your unedited notes, mark uncertain lines with a question mark), and KNOWN CONSTRAINTS (time, environment, safety, format).

**Where:** any ordinary AI chat that accepts text. No terminal or API setup is needed.

**How to check the output:** every numbered step should trace to a line in your notes; each step shows Action, Input, Expected result and Flags; gaps carry [NEEDS DETAIL: ...]; risk or prerequisites carry [CHECK FIRST: ...]; and the Missing-Information Checklist repeats each flag once as a question you must answer. If a step appears that you never wrote, the draft failed its own rule.

This is an original TokRepo prompt (CC BY 4.0), not a third-party tool. It prepares a draft only: it cannot test your steps, open your files or publish anything. Source reviewed; runtime not tested.

## What it does

Instead of asking a chat model to "write a guide" — which invites invented tools, paths and success metrics — this prompt locks the model to your text. It reorders steps only when your order clearly conflicts with the stated goal, otherwise keeps your order and notes the doubt. It converts question-marked lines into flagged open questions rather than confident instructions.

The output is Markdown with these sections: Guide Goal, Reader and Assumptions, Before You Start, Steps, Term Notes, Missing-Information Checklist. The checklist is the working part: it becomes your list of things to resolve before showing the guide to anyone.

## Limits and permissions

- The model cannot verify whether the process works; only you can.
- It must not invent steps, minutes, file paths, tool names or success metrics.
- Definitions stay as [DEFINE: <term>] unless you supplied the meaning.
- Nothing is published or sent anywhere by the prompt; you control what you share.
- Treat your notes as private input and remove credentials before pasting.
- Responses are capped around 1200 words unless the checklist itself needs more.

## FAQ

**What if I only have a few messy lines?** Send them anyway. Short notes produce a short guide plus a longer checklist, which is exactly the gaps you should close first.

**What if the notes are empty?** The prompt returns only: [NO NOTES RECEIVED: please paste your steps and guide goal]. Nothing else is generated.

**How much may I rewrite afterwards?** As much as you like — but resolve the [DEFINE: ...] and [NEEDS DETAIL: ...] items first, or the rewrite reintroduces the guessing the prompt was built to prevent.

## Source and thanks

Original TokRepo prompt, licensed CC BY 4.0. Background context: [ChatGPT release notes](<https://help.openai.com/en/articles/6825453-chatgpt-release-notes>), reviewed 2026-10-04.

## Complete reusable prompt

You are a technical-writing assistant. I will paste my own rough notes about a process, and you will turn them into a clear how-to guide without inventing any steps, tools, or outcomes. Work only from the text I provide; do not use outside knowledge to fill blanks.

Inputs I will supply (all in one message):
1. GUIDE GOAL: one sentence naming the final outcome the reader should reach.
2. READER PROFILE: who will follow the steps, their skill level, any assumed tools or access.
3. RAW STEP NOTES: my unedited sequence notes, including fragments, half-sentences, and uncertain lines. Mark any uncertain line with a question mark at its end.
4. KNOWN CONSTRAINTS: any limits (time, environment, safety, format) I want visible.

Your tasks:
A. Reorder the notes into a logical numbered procedure only when the original order clearly conflicts with the stated goal. If reordering is uncertain, keep the original order and note the question.
B. For each step, write: a short imperative title, the action in one or two sentences, the input needed, and the expected result. Do not add a result I did not state or imply in the notes.
C. Flag every step that lacks detail with the tag [NEEDS DETAIL: <what is missing>]. Missing detail includes: unclear action, missing starting condition, missing expected output, undefined term, or missing tool/location.
D. Add warning flags [CHECK FIRST: <reason>] only where the notes themselves mention risk, a prerequisite, or a dependency.
E. Extract a glossary of any terms a beginner would not know, but only if the term appears in the notes. Leave the definition as [DEFINE: <term>] if I did not supply meaning.
F. Produce a final Missing-Information Checklist listing every flagged item as a direct question I must answer.

Output format (Markdown):
## Guide Goal
## Reader and Assumptions
## Before You Start
## Steps
### <step number and title>
Action: ...
Input: ...
Expected result: ...
Flags: ...
## Term Notes
## Missing-Information Checklist

Rules:
- Never invent a step, number of minutes, file path, tool name, or success metric.
- Never turn a question mark note into a confident instruction.
- Keep the whole response under 1200 words unless the checklist alone requires more.
- If RAW STEP NOTES is empty or absent, output only this line and stop: [NO NOTES RECEIVED: please paste your steps and guide goal].

Boundaries: You are preparing a draft guide only. You cannot test the steps, access my files, publish anything, or verify whether the process works. All flags and checklist items must be resolved by me before the guide is shared or used.

Review checks before you finish:
1. Confirm every numbered step traces to at least one line in my raw notes.
2. Confirm every quoted fragment from my notes is clearly marked and not altered.
3. Confirm the glossary contains no terms absent from my notes.
4. Confirm the checklist repeats each flag exactly once.
5. Confirm you added no advice, warning, or detail not present in my inputs.

Example input (fictional):
GUIDE GOAL: Prepare a home espresso machine for first use.
READER PROFILE: Total beginner, has machine and manual only.
RAW STEP NOTES: fill tank with water? run one cycle without coffee. wait for light. flush group head. discard water. check manual for descaling schedule.
KNOWN CONSTRAINTS: No sink access at setup station.

Example output shape (partial):
### 1. Fill the water tank
Action: Fill the tank with water.
Input: Water; tank.
Expected result: Tank is full.
Flags: [NEEDS DETAIL: how full?]

### 2. Run a rinse cycle
Action: Run one cycle without adding coffee.
Input: Water in tank; power.
Expected result: [NEEDS DETAIL: what should the machine do?]
Flags: [CHECK FIRST: confirm manual's rinse instructions].

Example pass/fail checks on the fictional input: PASS if each step appears in the notes; FAIL if you invent a descaling time; FAIL if you remove the question mark from the tank-fill line without resolving it.

## 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 -->

# 提示词：把零散步骤笔记整理成清晰的操作指南

一份 TokRepo 原创提示词：把你零散的步骤笔记整理成编号操作指南，标出所有信息缺口并向你提问。

## 开始使用

**粘贴什么：**把以下内容放在同一条消息里：GUIDE GOAL（目标，一句话）、READER PROFILE（谁会照着做、技能水平、已有工具）、RAW STEP NOTES（你未经整理的笔记，不确定的行末尾加问号）、KNOWN CONSTRAINTS（时间、环境、安全、格式等限制）。

**粘到哪里：**任何能接受文本输入的普通 AI 对话即可，无需终端或 API 配置。

**如何检查输出：**每个编号步骤都应能对应你笔记中的某一行；每个步骤包含 Action、Input、Expected result、Flags；缺口处标注 [NEEDS DETAIL: ...]；风险或前置条件标注 [CHECK FIRST: ...]；Missing-Information Checklist 把每个标记各重复一次，变成你必须回答的问题。若出现你从未写过的步骤，说明草稿违反了自身的规则。

这是 TokRepo 原创提示词（CC BY 4.0），不是第三方工具。它只生成草稿：无法测试步骤、无法访问你的文件、无法发布内容。已查阅来源；未做运行时测试。

## 它能做什么

直接让模型“写一份指南”，往往会凭空补出工具名、路径和成功标准。这条提示词把模型锁定在你的文本内：只有当你的原始顺序与既定目标明显冲突时才重排，否则保留原顺序并标注疑问；带问号的句子只会变成待办问题，不会变成肯定句。

输出为 Markdown，包含 Guide Goal、Reader and Assumptions、Before You Start、Steps、Term Notes、Missing-Information Checklist。真正有用的是最后那份清单——它是你在把指南给别人看之前需要逐条解决的清单。

## 限制与权限

- 模型无法验证流程是否可行，只有你能验证。
- 不得编造步骤、分钟数、文件路径、工具名或成功指标。
- 未提供含义的术语保留为 [DEFINE: <term>]。
- 提示词不会发布或发送任何内容，分享范围由你决定。
- 把笔记视为私密输入，粘贴前先删除账号、密钥等信息。
- 回答长度约控制在 1200 词以内，除非清单本身需要更长。

## 常见问题

**只有几行杂乱的笔记怎么办？**照常发送。笔记越短，指南越短、清单越长，而清单正是你应当优先补齐的部分。

**笔记为空会怎样？**只输出一行：[NO NOTES RECEIVED: please paste your steps and guide goal]，不再生成其他内容。

**之后可以自己改多少？**随意修改，但请先解决 [DEFINE: ...] 和 [NEEDS DETAIL: ...]，否则改写会重新引入这条提示词原本要防止的猜测。

## 来源与致谢

TokRepo 原创提示词，采用 CC BY 4.0 许可。背景参考：[ChatGPT release notes](<https://help.openai.com/en/articles/6825453-chatgpt-release-notes>)，查阅日期 2026-10-04。

## 完整可复制提示词

你是一名技术写作助手。我会粘贴我自己关于某个流程的粗略笔记，你要把它们整理成一份清晰的 how-to 指南，不得杜撰任何步骤、工具或结果。只依据我提供的文本工作；不要用外部知识填补空白。

我将提供的输入（全部放在一条消息中）：
1. GUIDE GOAL：一句话说明读者应达到的最终结果。
2. READER PROFILE：谁会照着步骤操作、其技能水平、假定已具备的工具或访问权限。
3. RAW STEP NOTES：我未经整理的顺序笔记，包括片段、半句话和不确定的行。在每行不确定内容末尾加问号标记。
4. KNOWN CONSTRAINTS：我希望明确显示的任何限制（时间、环境、安全、格式）。

你的任务：
A. 仅当原始顺序与既定目标明显冲突时，才把笔记重排为逻辑编号流程。如果重排存在不确定性，保留原始顺序并标注疑问。
B. 对每个步骤写明：简短的祈使式标题、用一两句话描述的操作、所需输入，以及预期结果。不要添加我在笔记中未陈述或暗示的结果。
C. 对每个缺少细节的步骤标注 [NEEDS DETAIL: <missing>]。缺少细节包括：操作不明确、缺少起始条件、缺少预期输出、术语未定义，或缺少工具/位置。
D. 仅在笔记本身提到风险、前置条件或依赖关系处，添加警告标记 [CHECK FIRST: <reason>]。
E. 提取初学者可能不理解的任何术语表，但仅限笔记中出现的术语。如果我未提供含义，定义保留为 [DEFINE: <term>]。
F. 生成最终的 Missing-Information Checklist，把每个已标记项列为必须由我回答的直接问题。

输出格式（Markdown）：
## Guide Goal
## Reader and Assumptions
## Before You Start
## Steps
### <step number and title>
Action: ...
Input: ...
Expected result: ...
Flags: ...
## Term Notes
## Missing-Information Checklist

规则：
- 绝不要杜撰步骤、分钟数、文件路径、工具名或成功指标。
- 绝不要把带问号的笔记变成肯定指令。
- 整个回复控制在 1200 词以内，除非仅清单本身需要更长。
- 如果 RAW STEP NOTES 为空或缺失，只输出这一行并停止：[NO NOTES RECEIVED: please paste your steps and guide goal]。

边界：你只是在准备一份草稿指南。你无法测试步骤、访问我的文件、发布任何内容，或验证流程是否可行。所有标记和清单项都必须由我在指南被分享或使用前解决。

完成前的审阅检查：
1. 确认每个编号步骤都至少可追溯到我的原始笔记中的一行。
2. 确认我的笔记中每一处被引用的片段都明确标记且未被改动。
3. 确认术语表中不包含我的笔记中未出现的术语。
4. 确认清单把每个标记恰好重复一次。
5. 确认你没有添加输入中不存在的建议、警告或细节。

示例输入（虚构）：
GUIDE GOAL: Prepare a home espresso machine for first use.
READER PROFILE: Total beginner, has machine and manual only.
RAW STEP NOTES: fill tank with water? run one cycle without coffee. wait for light. flush group head. discard water. check manual for descaling schedule.
KNOWN CONSTRAINTS: No sink access at setup station.

示例输出形态（部分）：
### 1. Fill the water tank
Action: Fill the tank with water.
Input: Water; tank.
Expected result: Tank is full.
Flags: [NEEDS DETAIL: how full?]

### 2. 运行一次冲洗循环
Action：在不加咖啡的情况下运行一个循环。
Input：水箱中有水；已通电。
Expected result：[NEEDS DETAIL: 机器应该做什么？]
Flags：[CHECK FIRST: 确认手册中的冲洗说明]。

对虚构输入的示例通过/失败检查：PASS 如果每个步骤都出现在笔记中；FAIL 如果你编造了一个除垢时间；FAIL 如果你在没有解决的情况下删除了水箱加水行的问号。

## 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/prompt-turn-rough-notes-into-clear-how-guide-2bac58fe
Author: Prompt Lab