keymesh

Prompting AI Agents for `attogram/base` Projects

When using AI agents to create or update a repository based on attogram/base, providing a clear, upfront context is the most effective way to get the results you want. This guide provides two distinct prompt examples to help you steer AI agents in the right direction, whether you want to build upon the base philosophy or intentionally modify it.

Prompt 1: Using `base` As-Is

This prompt is ideal when you want to create a new project that adheres to the core, language-agnostic philosophy of the attogram/base template.

Example Prompt: Create a Personal Blog

This project builds upon the attogram/base template. Please integrate the new website content without removing the core base repository files or structure.

The main homepage for the site is `README.md`, and all new content should be linked from there. You can create new directories for your content as needed, but please do not modify the core `base` files and directories (such as `.github`, `docs`, `docker`, etc.).

Please use the existing Jekyll setup for the site, which builds from the root directory. Do not install other static site generators like Eleventy.

Finally, please look for and follow any instructions in an `AGENTS.md` file, as it may contain project-specific guidelines for AI agents.

Prompt 2: Modifying `base` for a New Technology

This prompt is for when you want to use attogram/base as a starting point but intend to introduce a specific programming language or framework. This example uses PHP and Laravel.

Example Prompt: Add a Laravel Backend

This project will use `attogram/base` as a foundation, but we will be modifying it to support a PHP/Laravel application.

Here are the requirements:
 **Add PHP and Laravel:** Please modify the Docker environment to include PHP and the necessary extensions for a Laravel application.
 **Install Laravel:** Once the environment is updated, install a fresh copy of the Laravel framework in a `/src` directory at the repository root.
 **Update NGINX:** Configure the NGINX server to point to the `public` directory of the Laravel application.
 **Preserve Core Files:** Do not remove the existing GitHub Actions workflows or the documentation in the `/docs` directory. The goal is to integrate Laravel into the `base` template, not replace it.
 **Check for `AGENTS.md`:** Please look for and follow any instructions in an `AGENTS.md` file, as it may contain project-specific guidelines for AI agents.

Why These Prompts Work

Providing a detailed, upfront prompt like the examples above is crucial for several reasons:

Sets Clear Expectations: The AI immediately understands the scope of the project. It knows whether to preserve the base philosophy or to modify it, which is the most critical distinction.
Reduces Errors and Rework: By specifying the technology stack (or lack thereof) and the desired file structure, you prevent the AI from making incorrect assumptions that would need to be corrected later.
Empowers the AI: A clear prompt gives the AI the constraints it needs to be creative and effective. It can focus on the “how” because you’ve already defined the “what” and “why.”
Faster, Better Results: Ultimately, a few minutes spent crafting a detailed prompt will save you hours of back-and-forth, leading to a better final product that meets your requirements from the start.

This site is open source. Improve this page.