LLM Prompt Engineering Guide

What's Inside

• Token economics & prompt stack — understand how models process your prompts
• System prompt design — production templates with role, constraints, output format
• Prompting strategies — zero-shot, few-shot, chain-of-thought with best practices
• Structured output patterns — JSON, XML, markdown tables, constrained output
• RAG prompt engineering — basic, citation-based, multi-hop, and follow-up generation
• Prompt chaining — extraction → analysis → report pipelines, agent loops
• Evaluation & testing — eval sets, LLM-as-judge, A/B testing prompts
• 30+ real-world templates — security analysis, code review, content, homelab, debugging
• Model-specific optimization — Claude vs GPT-4 vs Gemini vs local models
• Common pitfalls & anti-patterns — what not to do and how to fix it

Sample: Production System Prompt Template

You are a senior security analyst specializing in AI/ML 
infrastructure security. You have 15+ years of experience.

Your objective: Analyze security findings and provide actionable 
remediation guidance.

Constraints:
- Never speculate about vulnerabilities you cannot confirm
- Always cite the specific evidence from the provided context
- Prioritize findings by CVSS score and business impact

Output format: Markdown with sections:
  ## Summary
  ## Findings (table: ID | Severity | Description | Remediation)
  ## Recommendations (prioritized list)

Sample: Few-Shot Classification

Classify the security finding:

Example 1:
Finding: "SQL injection in login form parameter 'username'"
Classification: Critical
Reason: Direct database access via user input

Example 2:
Finding: "Missing HSTS header on web server"
Classification: Low
Reason: Defense in depth, not directly exploitable

Now classify:
Finding: "[YOUR FINDING HERE]"

Quick Reference Card

The guide includes a printable reference card covering temperature settings, when to use each technique, and model-specific tips for Claude, GPT-4, Gemini, and local models like Llama and Mistral.

Who This Guide Is For

• Developers building LLM-powered applications
• Security analysts using AI for threat intelligence
• Homelab operators running local LLMs
• Anyone who wants more reliable, consistent LLM output