nebunebu

ADVICE

3 min read

Advice for Refactoring and Publishing AI Prompts

To prepare your collection of AI prompts for a public site, I recommend focusing on categorization, standardized formatting, and descriptive naming. A public collection is most useful when users can quickly understand what a prompt does and why they should use it.

Group your prompts by their primary function. This helps visitors navigate the collection based on their immediate needs.

/prompts
├── learning/               # Education & Study tools
│   ├── atomic-flashcards.md
│   ├── anki-tutor.md
│   └── video-transcript-analyst.md
├── engineering/            # Coding & Dev tools
│   ├── codebase-auditor.md
│   └── cli-project-scaffold.md
├── research/               # Information Extraction & Analysis
│   ├── deep-text-extractor.md
│   └── concise-fact-extractor.md
└── index.md                # Main Table of Contents

2. File Refactoring & Cleanup

I identified some overlap and duplication that should be resolved before publishing.

A. The project_initialization.md Issue

Current State: This file is a mix of three things:

  1. Links to other files (index-like behavior).
  2. Two specific Anki prompts (“Prompt A” and “Prompt B”).
  3. A full duplicate of the code_audit.md text.

Recommendation:

  1. Extract the Anki Prompts: Move “Prompt A” and “Prompt B” into a new file named anki-tutor.md (or merge them into atomic-flashcards.md as “Alternative Strategies”).
  2. Remove the Duplicate: Delete the “Comprehensive Codebase Audit Request” section from this file, as it already exists in code_audit.md.
  3. Delete the File: Once extracted, delete project_initialization.md as it serves no distinct purpose.

B. Consolidating Extraction Prompts

Current State: You have extraction_prompt.md, extraction_prompt_2.md, and extraction_prompt_concise.md.

Recommendation:

  • Keep extraction_prompt_2.md as your “Flagship” extraction prompt. It is the most robust and well-structured. Rename it to deep-text-extractor.md.
  • Keep extraction_prompt_concise.md for users who want speed over depth. Rename it to concise-fact-extractor.md.
  • Archive or delete extraction_prompt.md if it is just an older version of version 2. If it has unique value, merge it as a “Variant” into the main file.

3. Naming & Titles

Good titles are “Action-Oriented” + “Noun”. They should describe the persona the AI adopts or the product it produces.

Current FilenameRecommended FilenameRecommended Public Title
card_creation_prompt.mdatomic-flashcards.mdAtomic Flashcard Generator
code_audit.mdcodebase-auditor.mdComprehensive Codebase Auditor
extraction_prompt_2.mddeep-text-extractor.mdDeep Research Text Extractor
extraction_prompt_concise.mdconcise-fact-extractor.mdConcise Fact Miner
project_setup.mdcli-project-scaffold.mdCLI Project Scaffolder (Nix/Claude)
yt_transcript_analysis.mdvideo-transcript-analyst.mdYouTube Transcript Analyst
(From project_initialization)anki-style-guide.mdAnki Card Style Guide

4. Frontmatter & Metadata

For a public site (especially if using Jekyll, Hugo, Docusaurus, or Next.js), add YAML frontmatter to the top of every .md file.

Example Template:

---
title: "Deep Research Text Extractor"
slug: deep-text-extractor
description: "A rigorous protocol for extracting every claim, fact, and citation from dense source texts without summarization."
tags: [research, analysis, writing, productivity]
complexity: Advanced
author: "Nebu"
date: 2025-12-17
---
 
# Deep Research Text Extractor
 
... content ...

5. Next Steps

  1. Split the content in project_initialization.md.
  2. Rename files to use kebab-case (lowercase with hyphens) for better web URLs.
  3. Add Headers (Frontmatter) to each file with a clear description.
  4. Create a central README.md or index.md that links to these categorized sections.

Recently Updated

Graph View