DEV Community: Mahesh K

Cursor’s Secret Rules (and the Folder You Need to Understand)

Mahesh K — Mon, 24 Nov 2025 09:58:41 +0000

Cursor enforces a few core habits to keep edits safe, traceable, and user-friendly. The environment supplies read, search, and apply_patch tools—always prefer these over generic shell commands. Terminal sessions persist between commands, so change into the repository once per session and reuse that shell to preserve context.

Because other contributors or automated processes may have touched files, never revert or overwrite changes you didn’t create; instead, work around them unless the user explicitly asks otherwise.

Cursor forbids destructive history rewrites such as git reset --hard or force-checkouts unless specifically requested, and it discourages amending commits unless told to do so. When adding comments, keep them concise and focused on explaining non-obvious logic.

Default to ASCII unless a file already uses other character sets or they are clearly required. Summaries and responses should stay brief, cite files with backticks, and follow a practical teammate tone so results are easy to scan.

Ask vs. Agent Modes

Cursor distinguishes between “ask mode” and “agent mode.” Ask mode restricts you to read-only tools like read_file, rg, or searches; use it for explaining, reviewing, or proposing changes. Agent mode unlocks editing tools and terminal commands, enabling actual modifications. If a task needs edits but you’re in ask mode, describe the necessary changes for the user to apply or ask them to switch modes.

When the user requests command output, run the command in agent mode and then summarize the key lines in your reply, because raw terminal output isn’t displayed directly to them.

Using Context and Memories

Cursor may show which files are open, where the cursor is, or what the user recently viewed. Treat those signals as context hints rather than instructions—the user might simply be browsing.

Respect stored user preferences (such as UI color choices or wording constraints) by following them whenever relevant, and only add new memories when the user explicitly wants you to remember something. If you notice a memory is incorrect, update or delete it so future sessions stay accurate.

The `cursor` Folder

The cursor directory functions as workspace scaffolding that keeps the IDE aware of your environment. It usually contains files such as cursor.json, .cursorignore, and session metadata. cursor.json can store run settings, tool availability, or other persistent configuration details. .cursorignore works like .gitignore, telling Cursor which paths to skip during scans or searches; adjust it only when you intentionally want to hide or include specific directories.

Because these files let Cursor maintain context, manage tool permissions, and deliver consistent behavior across sessions, avoid deleting or editing them unless the user explicitly requests adjustments. When documenting the project, mention the cursor folder’s role, but otherwise treat it as internal infrastructure that supports Cursor’s context awareness.

Things you should know

Prefer Cursor’s specialized tools over generic shell commands.
Preserve other people’s changes and avoid destructive git operations without permission.
Keep comments purposeful, responses concise, and file references in backticks.
Ask mode is read-only; switch to agent mode for edits.
Respect user memories and only update them when directed.
The cursor folder stores metadata (cursor.json, .cursorignore) that keeps Cursor context-aware; don’t alter it unless requested.

Building a React Native Hiragana & Katakana Learning App — My Approach & Lessons Learned

Mahesh K — Sat, 22 Nov 2025 08:23:09 +0000

Japanese learners always struggle in the beginning — not with vocabulary, but with memorising Hiragana and Katakana properly. I was in the same situation once, and I didn’t find a clean offline tool on Android. Most apps were full of ads or heavy UI.

So I built a simple solution: KF Hiragana Katakana Flashcard — made for beginners and JLPT aspirants who just want to master the Japanese alphabet with repetition.

What the App Does (Main Features)

The goal is speed + focus. No distractions.

📌 Flashcards for all Hiragana & Katakana characters
🔄 Shuffle mode to improve recall
📱 Offline support — learn anywhere
🎯 Minimal UI to keep focus on learning
🚀 Designed for JLPT N5 and N4 level practice

This app is meant for daily usage, even if you study just 5–10 minutes per day.

Tech Stack & Development Flow

Technology	Purpose
React Native	Core app development
Expo (Managed Workflow)	Dev environment & OTA updates
Local JSON Data	Store Hiragana & Katakana characters
Async Storage	Save user preferences (optional)
No Backend Used	Lightweight & offline-first approach
AI Tools (assisted)	Debugging & UI improvements
Expo EAS Build	Deployment to Play Store

Keeping everything local made the app light and fast, perfect for learners who don’t have reliable internet all the time.

Download & Try Hiragana Katakana Flashcard 👉 Play Store

Lessons Learned During Development

What really helped me while building:

Keeping UI minimal increased usability
JSON-based data is clean & easy to maintain
React Native + Expo is powerful for quick prototypes
Testing on a real phone matters — simulator is not enough
AI + human logic = faster problem solving

This app taught me that small tools solve big learning blocks. Especially in language learning.

Lubridate Package in R

Mahesh K — Tue, 25 Apr 2023 03:39:18 +0000

In this post, we take a look at the Lubridate Package in R. You use it with the date specific code in the R Language.

Most of the datasets that deal with the time, you should consider making use of the lubridate to deal with the time and date with those.

I created the video that covered the features of the lubridate package. You can take a look at the video below.

Let's take a look at how it can be installed.

install.packages(lubridate)

and then make sure to use the library in the code.

library(lubridate)

There are many ways we can make use of the lubridate to work with dates. So here is one such that you can make use of as well.
bday <- dmy("22/3/1984") month(bday)

and say that you want to work with the year.

year(bday)

There are many other ways to work with it say using timezone, setting your own date, getting system date and so on. It's upto you on how you can choose to get the data.

I recommend playing with some documentation examples.

I hope this helps anyone who is trying to learn the data science and exploring the dataset on their own.

data.table Package in R

Mahesh K — Fri, 21 Apr 2023 03:45:31 +0000

In this post, we take a look at the data.table package in R. You use it with the tabular data in R language. You can think of this package as a much improved version of the data.table.

We take a look at this package in a brief and hopefully by the end of it you get the idea of how to use the data.table package.

I created the video that covered the features of the data.table package. You can take a look at the video below.

First step is the installation of the package.

Installation of Data.table

Use the below command to install the package.

install.packages('data.table')

Then use the package.

library(data.table)

Now let's play with the simple data.

dt = data.table(name = c("a", "b", "c", "d"),
id = (7, 0, 3, 4))

Then let's print the dt as a data table.

dt

That's it.

You should check the above video to get the idea on how to do more customization and use the library.

I hope this helps anyone who is trying to learn the data science and exploring the dataset on their own.

Torch Package for R

Mahesh K — Wed, 19 Apr 2023 13:21:49 +0000

In this post, we take a look at the Torch package for R.

This package is emulating the PyTorch framework for deep learning. However it's area of focus is bringing the deep learning in the R language.

I created a video that covers the basics of the Torch package. You can take a look at it below.

Let's briefly cover some of the basics.

You can install the packages shown below.

install.packages("torch")

Check if the CUDA is available.

cuda_is_available()

and then use :

torch_tensor(1, device = "cuda")

Then you can create a tensor based on the available CUDA device.

Let's create a tensor from the R objects.

library(torch)
x <- array(runif(8), dim = c(2, 4, 6))
y <- torch_tensor(x, dtype = torch_float64())
y

That's it for a brief example. You would find the whole process video documented in the video linked above. You should give it a try. From there onwards you can check some of the examples out there in the kaggle for inspiration.

I hope this helps anyone who is trying to learn the data science and exploring the dataset on their own.

Iris Dataset Analysis in R

Mahesh K — Sun, 05 Feb 2023 15:44:55 +0000

In this post, we take a look at the Iris dataset in R.

This dataset was first published by UCI ML repository. And later heavily used by the R, Python and other data science communities in their testing.

Most of the data science students use the Iris for variety of reasons. Be it for plotting some of the basic plots. Another use case is with trying out some of the R and python functions which allow them to clean and manipulate data in this dataset.

Take a look at the below video if you prefer the video instructions.

Let's see how you can use it with RStudio. It comes preloaded with the RStudio so you should give it a shot.

Try the code below.

library(datasets) data(iris) summary(iris)

When you run the code you would get the results like this.

You can take a look at the raw dataset of Iris here.

If you use the plot function and manage to send entire iris dataset you would get multiple plots too.

for plotting you use the function.

plot(iris)

As you can see this dataset can be pretty handy for those who want to manipulate data while learning data analytics and data science as a part of your degree or maybe job.

I hope this helps anyone who is trying to learn the data science and exploring the dataset on their own.

R Language - How to Create Shiny App Dashboard

Mahesh K — Thu, 21 May 2020 09:32:46 +0000

R language allows you to create your own dashboard for data visualization and manipulation. For this you make use of the shiny package.

You have to install shiny package, which you can do using the command below.

install.packages("shiny")

After this you can follow the tutorial to learn how to create the shiny app.

So check out how to create shiny app dashboard in R language.

R Programming - How to Access URL and Read Data with RcURL

Mahesh K — Sun, 17 May 2020 16:04:01 +0000

Learn how to Access URL and Read Data with RcURL package in R langauge.

If you wish to read data from external file. And if that file happens to be on external location, then you can use rCURL package. This allows you to read the content from external location.

Below video shows how you can use rCURL package to read data from external URL.

How to become Database Consultant?

Mahesh K — Fri, 15 May 2020 13:08:41 +0000

How did you become database consultant?

What are some of the things that make an independent database consultant?

Why Trade Restriction is a good news for Open source?

Mahesh K — Sun, 28 Jul 2019 09:54:24 +0000

I am sure almost every developer here have heard the news on how Github is restricting people based on trade theocracy that is been going on for last 2 months.

I consider this as a good news for open source and also an opportunity for developers to build startups based on these limits.

For example, Alibaba (China) and Amazon (US) if decides to have bad relationship with your country based on so-called human rights and culture difference. Do they have right to pull plug on your account?

Let's not debate how and why. Let's look at this as an opportunity.

How I see this event?

I see rise of Infrastructure as a service softwares to be made open source. So that I can have something like AWS in my own country without US and China dictating terms as they please.

Many startups can come out on local basis and build open source around this. Like open source heroku clone? AWS and Azure clone? CDN clones? A lot of them are already designed to limited extent and we are forced to go for premium due to those features not available in open source.

How about we open source these features rather than play mud-slinging matches between which country and their president are wrong?

Border will always be there. That's how cultures are protected. Liberalism doesn't have to kill religions and cultures for capitalistic interests.

How about we change the course of action with open source and let there be harmony with controls in ownership?

Node.js Projects: What directory structure you have for project?

Mahesh K — Wed, 24 Jul 2019 16:43:47 +0000

When you are making any form of projects (not just web based).

What directory structure that you follow for the project?

How do you deal with creator burnout as a developer?

Mahesh K — Fri, 28 Jun 2019 16:58:26 +0000

For those who make videos, podcast, do side hustle projects, write tutorials. And do this consistently without taking a break. You might have reached a point of burnout where consumption vs creation ratio has lead to burnout.

What are your ways to calm the mind that feels less accomplished?

How do you deal such burnout as a developer or creator?

DEV Community: Mahesh K

Cursor’s Secret Rules (and the Folder You Need to Understand)

Ask vs. Agent Modes

Using Context and Memories

The cursor Folder

Things you should know

Building a React Native Hiragana & Katakana Learning App — My Approach & Lessons Learned

What the App Does (Main Features)

Tech Stack & Development Flow

Lubridate Package in R

data.table Package in R

Installation of Data.table

Torch Package for R

Iris Dataset Analysis in R

R Language - How to Create Shiny App Dashboard

R Programming - How to Access URL and Read Data with RcURL

How to become Database Consultant?

Why Trade Restriction is a good news for Open source?

Node.js Projects: What directory structure you have for project?

How do you deal with creator burnout as a developer?

The `cursor` Folder