prompts

Clipboard Health's collection of prompts to assist LLMs in generating code adhering to our best practices.

Table of contents

• Usage
• Directory structure
• Attribution

Usage

# Add this repo as a submodule to your project
git submodule add [email protected]:ClipboardHealth/prompts.git

# Create a symlink to an appropriate rules file in the current directory
ln -s ./prompts/rules/fullStack.md .cursorrules

When prompting LLMs, @ reference the prompts by name. For example, @_test.md @myFile.ts writes tests for myFile.ts. See @ Symbols for details.

Directory structure

• fragments/
  • Reusable prompt fragments that ./combine.sh concatentates into complete prompts or rules.
• prompts/
  • Provided to LLMs to generate best practice code. Start names with _ for easy @ references. See examples in Usage.
  • beta/
    • Ideally, you'll get reliable results before adding prompts to this repository, but sometimes it's nice to iterate across various domains. Use this folder for the latter.
• rules/
  • Custom instructions provided in every prompt. See a symlink example in Usage and Cursor's Rules for AI for details.

Attribution

• Our rules are a combination of those in Cursor Directory.
• _chainOfThought.md is from r/cursor.
• _commitMessage.md (which heavily inspired _prCreate.md) is from dgokcin/dotfiles.
• _proofread.md is from r/ChatGPTPromptGenius.
• _reasoning.md is from philschmid/prompt.txt.
• Anthropic's prompting courses.