---
title: "LLM Prompts | DJ Petersen"
description: "A curated library of LLM prompts for coding, debugging, architecture, writing, and more."
canonical: "https://www.thedjpetersen.com/llm-thoughts/prompts/"
---

A Prompt Library:

# Prompts

A curated collection of prompts for working with LLMs — for coding, debugging, architecture, writing, and more. Each prompt is a tool shaped by use.

[## Two Things That Make Code Maintainable Wise words from the best systems engineer I've worked with. Reduce the layers a reader has to trace and the state they have to hold in their head. Applies to every codebase. Always.](https://www.thedjpetersen.com/llm-thoughts/prompts/maintainable-code-two-things)[## The Prototype Interview Before writing a line of code, have the AI interview you about your vision. 50 rapid-fire questions extract design decisions and judgment you didn't know you had — then the model can one-shot a working prototype roughly aligned with the picture in your head. via Geoffrey Litt](https://www.thedjpetersen.com/llm-thoughts/prompts/prototype-interview-kickoff)[## Critique This Prototype After you've built something, have the AI tear it apart — not as a code reviewer, but as a product thinker. What assumption is each screen testing? Where does the vision leak? High fidelity can mask the absence of shared understanding.](https://www.thedjpetersen.com/llm-thoughts/prompts/prototype-critique)[## Prototype to Production The prototype works and the concept is validated — now harden it for production. This prompt walks through the unglamorous work of turning a vibe-coded draft into code you'd actually ship.](https://www.thedjpetersen.com/llm-thoughts/prompts/prototype-to-production)[## Give Me Five Directions When you're stuck on what to build or how to approach a problem, ask for five wildly different directions — not variations on a theme. Forces the model out of its clustering tendency and into genuinely divergent territory.](https://www.thedjpetersen.com/llm-thoughts/prompts/divergent-directions)[## The Careful Coder's Manifesto Operating principles for agentic code intelligence — measure twice, cut once.](https://www.thedjpetersen.com/llm-thoughts/prompts/careful-coders-manifesto)[## Code Smells from First Principles Why code smells exist, grounded in cognitive science and information theory.](https://www.thedjpetersen.com/llm-thoughts/prompts/code-smells-first-principles)[## Code Quality Auditor Framework Structured 3-phase audit prompt for identifying and eliminating quality-lowering patterns.](https://www.thedjpetersen.com/llm-thoughts/prompts/code-quality-auditor)[## Meeting Transcription Context Transform meeting transcriptions into publication-quality documentation with decisions, action items, and reasoning.](https://www.thedjpetersen.com/llm-thoughts/prompts/meeting-transcription)[## Edge Cases You Didn't Consider After plan mode, force the model to audit its own blind spots. It's silly that it works, but it does — asking the model to reflect on what it missed surfaces real gaps. via Brian Lovin](https://www.thedjpetersen.com/llm-thoughts/prompts/edge-cases-after-plan)[## Make It Simpler and Dumber Push back on over-engineering by asking for the dumbest possible solution that still achieves the goal. Models tend toward complexity — this resets the bar. via Simon Last](https://www.thedjpetersen.com/llm-thoughts/prompts/simpler-and-dumber)[## Capitalize for Emphasis Does capitalizing words for emphasis actually work, or is it placebo? Consensus leans toward: it works. CAPITALIZED words signal priority and weight to the model. via Brian Lovin](https://www.thedjpetersen.com/llm-thoughts/prompts/capitalization-for-emphasis)