Better Use AIs in VS Code with Prompt Files

Why Prompt Files Matter

At 0:00 Tim says, “Using a prompt file with an AI can turn the experience from just okay to great.” A prompt file tells the AI model how to think, how to proceed, and how to respond to a user prompt. Tim links this to Burke Holland’s “Beast Mode” file from Microsoft, a large system prompt that acts like a template. By using this, developers can cut down on repetitive instructions, reduce mistakes, and get more predictable generated code inside VS Code.

Tim frames it as a kind of prompt engineering: you’re building a reusable set of instructions so that when you later type a quick user prompt in the chat view or the AI Toolkit icon on the activity bar, the AI already knows your coding style and preferred project structure.

Tim’s Developer Resources

Before showing the tool, Tim (0:27) recaps the free and paid resources he offers—nine full courses, a tenth on iamtimcorey.com, plus a podcast answering developer questions. He highlights that these are meant to help developers refine their skills and build better workflows. This sets the stage for why a VS Code AI prompt file matters: it’s part of modern practices for improving everyday coding tasks.

Setting Up Beast Mode in VS Code

At 1:22 Tim moves into the tools section of VS Code. He already has Beast Mode installed and selected under “configure modes.” He explains that you can open the AI Toolkit view, browse models from the model catalog, and create a new custom mode. In his example, you might click Select Add Prompt, give it a name like “GitHub Chat Modes,” and paste in the file’s contents (1:34). This stores your custom system prompt so it’s always ready in your workspace.

Tim’s setup shows how local models or cloud models can be combined with your own prompt builder instructions. Everything sits neatly in your VS Code root folder or another stored folder, making it easy to repeat later.

Using AI on a Real Project

From 2:03 Tim demonstrates the AI on his actual website project hosted in his VS Code workspace. He opens the code for the “Learning Paths” section—essentially a Python/React-backed site—and shows the app navigation menu. He wants to tweak the arrow icon to add a small animation on hover (3:21).

This is a typical developer scenario: you have some front-end code, you’d like an AI agent to generate the CSS for you, and you don’t want to hand-write every detail. With a prompt file already loaded, your user prompt can be short, and the AI will fill in the context.

Preparing the AI Prompt

At 3:38 Tim uses the Windows Snipping Tool (Windows+Shift+S) to capture a screenshot showing the arrow (3:45). This visual context helps the AI. He then types a natural prompt at 4:43:

“In my navigation menu under the learning path section, rotate the arrow from 30° to horizontal when you mouse over it, then back on mouse off. Here is a screenshot of what it looks like currently.”

Tim notes he might later specify exactly where to put the CSS, but he begins with this following prompt to test the functionality of Beast Mode.

Beast Mode’s Plan

At 6:00 Tim explains that Beast Mode outlines a workflow: identify the correct selector, add a CSS class, add a hover effect, then test and verify. It even places the CSS in the correct site.css file above the root (6:42). This shows how a predefined schema in your prompt file—your “rules” about where styles go—guides the AI’s response.

Tim calls this the “big deal” of using a prompt file: it adds context, letting the AI operate as if it knows your project structure without you having to repeat the full description every time.

Iterating and Refining

When Tim checks the site at 6:54, the generated code only affects one arrow and in the wrong direction. He then refines the prompt at 7:29 to flip the hover effect and apply it to all arrows.

At 8:49 he notices that now all arrows rotate at once. So he sends another user prompt at 9:06 telling the AI to apply the effect only to the hovered item. Tim remarks at 9:38, “You have to tweak an AI. It doesn’t always do the right thing.”

This is a great example of prompt engineering inside the chat view. You edit, repeat, and save your prompts until the AI produces the structured output you want. Because the prompt file already supplies the system prompt, each new instruction is just a small refinement.

Final Result

By 9:52 Tim confirms the animation now works per arrow: rotate on hover, reset on mouse off. He notes you could switch to the light or switch to the dark theme, change colors, slow the transition—whatever you like (10:02). The key is that the AI figured out the SVG transform and wrote the CSS for him (10:11).

Tim explains at 10:30–11:20 that you can extend Beast Mode or create your own prompt builder file to define how you lay out files, how you want code formatting handled, or how to call a Python function. Then your day-to-day prompts can be much shorter and still yield consistent answers.

Reviewing AI Output

At 11:50 Tim warns to always review the AI’s changes. Even with a protected prompt file and good practices, it’s possible for an AI to add a class in the wrong place or break your required dependencies. In his example the AI made only small, sensible changes—adding a CSS class here, a line of code there—but developers should still check the log of changes and refine if needed.

Tim closes by noting that once you have a result you like, you can save it and reuse it. That’s the essence of using AI models and prompt files inside VS Code: a repeatable, consistent workflow that turns ad-hoc prompting into a reliable SDK-like experience.

Takeaways from Tim Corey’s Demo

Following Tim Corey’s walkthrough in the video shows how a well-designed prompt file inside Visual Studio Code can:

• Provide consistent context to AI models so your prompts can be shorter.

• Handle generated code, structured output, and formatting automatically.

• Integrate into your workspace, activity bar, and AI Toolkit so you can easily open, test, and refine prompts.

• Allow you to browse models, switch between local models or cloud, and apply your prompt engineering rules seamlessly.

By storing your rules in a prompt file, you give the AI a reliable agent script to follow, letting you focus on higher-level development tasks instead of repeating the same instructions.