AI Supports Project Management and Stress! A Practical Course on Cursor Agents for Non-Engineers

Hello, I'm Miyatti @ Explaza, a huge Cursor fan (my X account is @miyatti ).

In the previous article, "Cursor Agent Super Introduction", I introduced light examples of letting AI handle dinner preparations and daily tasks.

I thought this niche topic of a non-engineer using Cursor, an editor for engineers, would only appeal to a specific audience, but I was incredibly happy to see that it was read by a much wider range of people!!!

Yesterday, I gave an LT (which is also the basis for this article) as a follow-up at Findy's PM LT Night, and it was awesome to hear things like "I'm a PM, but I started using Cursor after reading your article!".

By the way, here are the presentation slides.

So, that's how it went, and I gave this super advanced version as an LT, but I couldn't really go into much detail during the LT, and I made it a "Can you make it understand your heart?"... no, that's not it edgy advanced version lol. So, I thought, "I should write a proper article with more realistic topics", and started writing this morning.

The theme this time is "Using 'Rules' to become a more autonomous agent".

Do you remember when I mentioned in the previous article that "to make subsequent tasks more efficient, it's best to eventually compile your prompts"?

Actually, Cursor has a feature called "Rule", where you can build multiple "agent modes" just for you. For example,

• Task Management

• Idea Generation

• Project Management

• Mental Support

and so on. Just by setting up Rules, Cursor will act as if it is switching "personalities" to work and support you autonomously.

In this article, we will dive deep into that "advanced version using Rules"!

1. Review of the basics: Why are "subsequent" prompts important?

In the previous article, I introduced how you can interact with AI by brainstorming loosely at the beginning, and finally

and when you ask that, a template like a “specification document” is automatically generated—that was the story I shared.

This “create once, use many times” template strategy was one of the strengths of the Cursor agent.

How did you do it until now?

Previously, even if you had a good prompt or a template flow,

It must have felt quite manual in some way.

The world after introducing Rules

However, when you introduce something called the Rule feature, Cursor will automatically or in response to specific keywords (mode switching), it will load that template.

• From a human perspective, you just have to say “Please do this in XX mode!”,

• AI begins to act autonomously, thinking, 'Ah, it's best to refer to the template written in this Rule!'autonomously

This is the reason it looks like 'autonomy'. By turning it into a

Rule, the AI will start proactively suggesting steps ahead for you.

For example, from preparing dinner to project management

Taking dinner preparation and daily tasks as examples, this mechanism that makes 'the second time onwards super easy' can also be applied to larger work—such as project management and team tasks.

Rule allows you to:

This establishes a workflow. Since you can
semi-automatically reuse what you wrote once, you can run things with almost zero effort from the second time onwards.

If you have mastered the flow of 'initial divergence → specification creation', let's use Rules to distill that into a form that can be called up by anyone at any time. This is the evolved version of the
Cursor Agent, and the gateway to ' autonomy through Rules,' which is the main theme of this article.

2. What is Cursor's 'Rule' feature?

'It's just a Markdown file, but the agent reads it autonomously'

Rule files are actually just text files containing simple prompts (Markdown format). The key point is that by describing 'what situations to use them in' and 'what content to include,' Cursor reads them and influences the AI's behavior.

• Reads them when necessary (automatic judgment or specified keywords)

• “Official documentation that is always referenced” which the AI can easily understand

Concrete Example

For example, let's imagine a scenario where you are planning dinner. If you create a Rule file named
dinner_planning and write down the flow of dinner preparation (check the fridge -> suggest a menu -> consider the budget...), you can just say 'Please go into dinner mode!' and Cursor will automatically guide you through menu planning.

Previously, you had to do the tedious work of 'keeping a note of menu templates you thought of before and copying and pasting them when you wanted to use them.'

After introducing Rules, Cursor itself autonomously loads that template and even “anticipates” and makes suggestions for you.

3. Basic creation and setting examples for Rules

Step 1: Add a task management Rule to Cursor's 'Project Rules'

Step 2 Write the Rule

For example, like this.

# 夕飯プランニングRule

## 目的
- 家族構成（子供2人と大人2人）を考慮した献立を提案
- 辛さ控えめ＆時短レシピを優先
- 予算を抑えたい

## 具体的な流れ
1. 冷蔵庫にある具材を確認（ユーザーに質問）
2. 料理の要望・制約をヒアリング
3. ベースの献立案を2～3パターン提案
4. タイムライン（買い物→下ごしらえ→調理）を生成
5. 仕上げ＆盛り付けヒント

## 注意点
- 辛味調整など、子供の嗜好に合わせたオプション提示
- 予算を超えそうなときは安価な代替具材を提案

If you write this, it will actually be saved to ~/.cursor/rules/dinner_planning.mdc. Once saved, it should be recognized automatically.

* You're curious about this part at the very top, right? Well, for now, don't worry about it and just set it to Always.

To explain this roughly

• Always: It will read this whenever you make a request (it tries its best, though sometimes it doesn't).

• Auto Attached: If you specify which directory or file to apply it to, it will only read this request within that scope.

• Agent Requested: The agent will autonomously read it when necessary.

• Manual: You manually mention or specify it to have it read.

That's the general idea.

Especially important are Always and AgentRequested. If you're not sure about the details, Always is fine. The best approach is to set it to AgentRequested and write a description of that mode in the Description field.

If you set it up like this, when you tell the Agent to "please use XXX mode," it will be able to read it correctly.

Actual usage: Switching to Dinner Mode

For example, you call it like this:

Then, it will proceed with automatic Q&A while keeping in mind the specific flow and points to note from the "Dinner Planning Rule" written earlier.

The AI will anticipate advice such as "adjusting the spiciness" or "what to do if you're over budget?".

Hands-on 1: Creating a Task Management Mode

Let's look at an example of how to actually write a Rule for a "Task Management Mode".

todo_management.mdc

# タスク管理モード（シンプル版）

## 目的
- TODOファイル (例: `todo.md`) を使った日常タスク管理をサポート
- 新しいタスクがあれば自動追記
- 完了報告を受けたらタスクを完了扱いに

## 手順イメージ
1. **TODOファイル存在確認**  
   - 無い場合 → ユーザーに「今やるべきタスクは？」と質問 → もらったタスクをもとに新規で`my_todo_files/todo.md`を作る  
   - ある場合 → 既存の`todo.md`を読み込み、今日やるべきタスクを優先順に提示
2. **タスクの追加**  
   - ユーザーが「新しいタスクがある」と言えば、TODOファイルに追記
3. **タスク完了報告**  
   - 「このタスク終わった」と言えば、そのタスクに完了マーク（例: `[x]`）を付ける
4. **タスクの相談・分解**  
   - ユーザーが「このタスクどうやったらいい？」と聞いたら、必要に応じてサブタスクへ分解 → TODOファイルに記載

## 注意点
- **優先度が高いもの**を上に（ユーザーへ確認しながら並べ替える）
- **締切がある場合**は期限をメモ
- 可能なら**タスクを短く分割**して追加

Just by writing this, the moment you say "Task Management Mode," Cursor will automatically create or reference todo.md and assemble the optimal task list.

Usage concept

It guides you by asking things like, "How did yesterday's tasks go?," "What is your top priority today?," or "Shall we do a daily review?."

Convenience points

• Your daily morning routine becomes easier. If you say, "
  Please enter today's task management mode," the AI will immediately show you today's schedule or suggest what to do in your spare time.

• The time humans spend worrying about "Wait, was I supposed to do this next?" is significantly reduced.

If you don't call this "convenient, what else would you call it...!

4. Supporting your own work: Changing your "persona" by switching modes

4.1 Creating Rules to match your work

Some might think, "Are Rules only for 'dinner ideas'?" but they actually show their true power in business scenarios. This is because daily work often involves flows that can be standardized. Task management isn't just for curry, after all lol

If you set each one as a Rule called "XX Mode," Cursor will behave as if it is switching its "persona" depending on the situation.

Example: What kind of modes should you prepare?

Task Focus Mode

• When performing a specific task, use Chain of Thought (CoT) to plan the steps.

• Example: Automatically guides you through processes like "First prepare materials -> Create draft -> Review -> Complete."

• Because the AI anticipates and suggests "what to do next," you can proceed efficiently.

Divergent Thinking Assist Mode

• When you need ideas, the AI leads divergent thinking techniques such as SCAMPER (a creative ideation method).

• “How about combining these?” or “What are other uses?” — because it asks concrete questions, the benefit is that you can brainstorm ideas even when you are alone.
  

The idea is that for all of these, you just say “Help me in XX mode,” and Cursor provides the optimal steps or questions. The great thing is that
once you create a Rule, you can use it repeatedly. In the second half, we will actually introduce how to create these modes.

5. Application to Project Management

Now, besides the above, from here on we will introduce several recommended examples of how to use these Rules even more conveniently in practical work.

“Can't Cursor be used not just for daily tasks and brainstorming, but for serious project management?”
Of course it can.

Moreover, by fully utilizing Cursor's Rule function, you can automatically organize background information for large-scale projects and document creation.

5.1 Making Background Information into Rules

5.1.1 Why is project background information important?

Large-scale projects have a vast amount of so-called “background information.” For example:

• Industry knowledge (market reports and competitor trends)

• Existing specifications and minutes (meeting memos, past documents)

• List of stakeholders and roles (who is the leader, who is the reviewer, etc.)

Honestly, it is tough to “explain to the AI” every time. It is a hassle to say one by one, “This time it's a medical project…” or “This person is the client…” or “Refer to this document…”

5.1.2 Image of “Project X Mode”

So, you just write the background information for Project X entirely into the Rule file.

1. Create a Rule file named projectX.mdc

2. Include items like "Project Overview," "Related Folder Structure," "Stakeholder Names," and "Regulations or Specifications to Note"

Then, once Cursor is told "I want to do this task for Project X",

It will automatically understand and reference everything at once, including a large amount of background material.

5.2 Actual File Generation Workflow

5.2.1 Define Rules at the Start

First, instruct in the Rule that "Requirement definitions should be summarized in docs/requirement.md".
Furthermore, write down things like "Specifications go in docs/spec.md, and meeting minutes go in the meeting_notes/ folder."
It is also great to list stakeholder lists or project priority criteria while you are at it.

Example: projectX.mdc

# プロジェクトXモード

## プロジェクト概要
- 大手保険会社向けのシステム開発
- スケジュールは来年3月末まで
- 厳しいセキュリティ基準あり

## ドキュメント保存先
- 要件定義書 → `docs/requirement.md`
- 仕様書 → `docs/spec.md`
- 議事録 → `meeting_notes/`
- タスク管理 → `todo/`

## ステークホルダー
- プロジェクトリーダー：山田さん
- セキュリティ顧問：鈴木さん
- 技術リーダー：田中さん

## 注意点
- 個人情報を含むデータはマスキング必須
- 顧客の承認フロー：山田さん→セキュリティ顧問→顧客

Set the Rule Type to Agent Requested, and if you write something like "description: A collection of rules for Project X. Specifies where to save requirement definitions, specifications, meeting minutes, etc.," it will read it when you bring up the topic of Project X. It will also read it if you specify it.

5.2.2 AI Automatically Writes to Files

After that, when you say "Add this to the Project X requirement definition document" in the conversation,

• Cursor thinks, "Oh, that was written in projectX.mdc. The requirement definition document is docs/requirement.md"

• Moreover, it references all files under projectX/ and pulls in necessary related materials

• The user no longer needs to specify "where to save what" one by one

Meeting minutes are the same. If you ask, "Add the meeting notes for Project X", it will search for and add meeting_notes/* for you.
As a result, we only need to give a few lines of instructions.

5.2.3 Reducing the hassle of manual copy-paste management

In the past,

• when you wrote new meeting minutes, you would move them to the meeting minutes folder

• when there was a change to the specifications, you would find the corresponding file yourself
  ... these were tedious tasks that you likely had to do often.

Rule allows you to automatically edit files (while adjusting the text with CoT if necessary), so the "Wait, where is that file?" type of confusion is drastically reduced.

In short, you can offload the "copy-paste labor" of document management to AI.

1. Just write the background information once

   • and the AI will always refer to it

   • The more complex the project, the greater the effect

2. Automatic document storage

   • By summarizing storage locations and notes in the Rules, there is almost no need for manual organization by humans

   • Consistency in file naming and folder structure is maintained

3. Meeting minutes, specifications, and task lists updated before you know it

   • Just by asking, "I want to do X in Project Y," the necessary files are edited automatically

   • Meeting notes and the like are also automatically generated according to a fixed format

The liberation from "copy-and-paste hell" is something those in charge of large-scale projects should be able to deeply relate to.

As a secondary effect, a folder structure that is easy for AI agents to understand is also easy for humans to understand. It is a great advantage that the folder structure remains easy to understand later, even when you create or save files manually yourself or share them with others.

6. Using it for mental support too? — Refresh with bird's-eye view mode

When you are busy with work, isn't there a moment when you think, "I want to cool down a bit"? Actually, by using Rules, you can design it so that even

rest and mental care are led by AI.

7.1 Resting is also a proper task

"Resting when you are tired is also a proper job". If you set things like
bird's-eye view mode as a Rule,

• "I'm feeling a bit rushed and need to calm down."

• "I'm feeling unfocused and overwhelmed."

When you feel this way, the AI will automatically prompt you with deep breathing guides or reflection sessions.

7.1.1 What kind of support can you receive?

1. KPT Reflection

   • Using Keep / Problem / Try to categorize "what went well today," "issues," and "things to try next"

   • allows you to objectively organize your actions and thoughts.

2. Mindfulness

   • Breathing techniques (carefully guiding how many seconds to inhale, hold, and exhale)

   • Simple stretches and other instructions suggested by AI

   • Isn't it interesting to have "AI-guided relaxation"?

3. Exercises to encourage positive thinking

   • "Let's list three good things from today"

   • "What are you grateful for?"

   • We have prepared mechanisms to help you feel a bit more positive, even on stressful days.

By setting these procedures as Rules for the AI, you can instantly start personalized mental care anytime just by declaring, "Go into bird's-eye view mode!".

7.2 Also for business improvement through reflection

Mental support is not just about 'being comforted'.
While reducing fatigue, it is also an opportunity to rethink your work methods and efficiency.

1. Reflection function

   • When you ask the AI to 'Let's look back at today's work and organize things a bit', it will identify issues and improvements from a Chain of Thought perspective

2. Improving work processes with KPT

   • You can review the work flow itself, which also leads to improved performance for the entire team

3. Regular cool-downs

   • When you keep working, you tend to just keep running. Even just having the AI suggest, 'Why don't you take a break?' can be a trigger to increase long-term productivity

Rule by systematizing regular breaks and reflections, a major benefit is that you can avoid “working too hard and burning out” and avoid missing important insights.

7.3 Let's try creating a Rule like this

For example, create a Rule file named mental_support.mdc,

• Bird's-eye view mode

• Stretch mode

• KPT mode

and group sub-modes like these together.
“I'm tired... please use bird's-eye view mode” → Deep breathing → Organizing thoughts → Simple TODO adjustment → Refresh complete, a sequence that can be set up in no time.

Conclusion: We are in an era where rest is also part of the work process, and AI can assist with it. If you create a Rule for mental support in
Cursor, you can feelings of impatience or stress and immediately “switch modes” to get a bird's-eye view of your own mind. Please try it to balance work and rest.

8. Summary──Just by writing a “Rule”, Cursor agents accelerate

So far, we have looked at how to utilize Rules for multiple modes. Once again, by introducing the
Rule feature, Cursor will act autonomously as if switching personalities to support you.

8.1 You can "switch personalities" by creating modes with Rules

1. Task Management Mode

   • Automatically organize TODO lists and suggest priorities

2. Task Execution Mode

   • Plan with Chain of Thought and track progress

3. Idea Generation Mode

   • Guide through brainstorming techniques (such as SCAMPER)

By creating as many of these modes as you like, you can dramatically streamline your work and life.

8.2 It can also be applied to project management

• Background information (industry knowledge, specifications, meeting minutes, stakeholder information) turned into Rules

• "Project X Mode"—just by saying this, the AI will automatically reference vast amounts of material, update documents, and create meeting minutes

• Since it eliminates the hassle of explaining from scratch every time, it's convenient because it instantly grasps and supports the overall project picture for you.

8.3 Mental support and overview mode are also available

It is useful not only for work but also for daily mental care.

• “I'm tired…” → AI prompts deep breathing guides and KPT reflections

• “I'm feeling rushed and want to calm down” → Overview support mode helps organize your thoughts

Life × Work, the Cursor agent supports you as your “personal secretary” in both aspects.

8.4 “Blazing fast from the second time onwards” — Rules are the shortest route

As I have already mentioned several times, the biggest advantage of the Rule feature is that “it is super smooth from the second time onwards”.

• First time: Trial and error to distill specifications and templates into Rules

• Second time onwards: Just by specifying the mode, the AI anticipates and prepares the optimal flow

“Once you write the rules, you can reuse them many times” directly leads to work efficiency and stress reduction.

Next steps: MCP integration & team sharing

This time, it was honestly too long lol, so I couldn't include the following content, but please give it a try! (I might write about it next time)

1. Other tool integration (MCP)

   • Integrate with external services like Slack, Gmail, and Google Calendar, and evolve to the level where “the AI secretary navigates the internet world”

   • Example: Summarize Slack conversation points, refer to the latest request emails in Gmail, and automatically enter schedules into the calendar... and so on

2. Share Rules across the entire team

   • Work can be carried out using a unified format and workflow

   • Even when new members join, they can refer to the same Rules to learn the flow immediately

Just by writing Rules, agents become 'more autonomous'

Cursor goes beyond being just a chat AI, it is something you can cultivate as a 'personal assistant'. The key to this is the
Rule feature. 'Blazing fast work,' 'preventing oversights,' 'optimizing the entire team,' and 'improving life balance'—all of these become much easier to achieve if you have
Rules.

Please try creating Rules that are perfect for your own work and life.
Just by writing 'Rules,' Cursor agents will become stronger and act more autonomously!

Why not take this opportunity to further power up your AI assistant?

(Updated 2025-05-18)
I wrote an article about how to write more concrete and accurate Rules. If you are interested, please take a look at this as well.

[Final Advertisement] Explaza is looking for new team members!!

Diverse positions from 'AI Engineer' to 'AI Producer'

Explaza, where I work, is in a rapid expansion phase and we are looking for colleagues to take on the following roles.

For detailed job requirements and benefits, please check our recruitment page.

• LLM Application Engineer (Full-time)

• AI Producer (Full-time)

• Lead AI Producer (Full-time)

• Open Position

We are especially looking for AI Producer (PdM) positions!

What kind of company is Explaza?

We are an implementation-focused consulting and product company that handles the social implementation of generative AI. Founded in 2020, we are still a young company with just over 10 members, but we are a notable startup that advocates for the 'social implementation of generative AI' using both AI technology and product development as our weapons.

Click here for our culture deck

In 2023, we formed a capital and business alliance with the Matsuo Institute, and from 2025 onwards, we are entering a phase of fully strengthening our generative AI capabilities!

At Explaza, we have made Cursor, which was featured in this article, available for unlimited use by everyone in the company, regardless of whether they are in business or engineering roles!

For everyone who became interested in using Cursor in their daily life through this article!Why not join us and help build an AI-native society!! You get unlimited use of Cursor too! lol We look forward to hearing from you!