Use when designing prompts for LLMs.
Scanned 9/10/2026
Install to Claude Code
npx -y skills add LoopyLuci/Skills --skill prompt-engineering-patterns --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Prompt Engineering Patterns?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/loopyluci-prompt-engineering-patterns)More formats (shields.io, HTML) on the badges page.
---
name: prompt-engineering-patterns
description: "Use when designing prompts for LLMs."
category: mlops
tags: [prompt-engineering, llm, gpt, patterns, prompting]
---
# Prompt Engineering Patterns
Systematic patterns for designing effective LLM prompts.
## Core Prompt Structure
```
System: [role + rules + constraints]
Context: [background information]
Examples: [few-shot demonstrations]
Input: [the actual query]
Output: [format specification]
```
## Zero-Shot
```
System: You are a Docker expert. Answer concisely with commands.
User: How do I remove all stopped containers?
```
## Few-Shot
```
System: Translate Docker commands to explanations.
Examples:
User: docker rm $(docker ps -aq)
Assistant: Removes all containers (running and stopped).
User: docker system prune -a -f --volumes
Assistant: Removes all unused containers, networks, images, and volumes.
User: docker compose up -d --build
```
## Chain-of-Thought (CoT)
```
System: Solve step by step.
User: I have a Docker container that exits immediately.
Think through this step by step:
1. The container needs a foreground process to stay alive
2. Check the CMD/ENTRYPOINT in the Dockerfile
3. For nginx, the default CMD runs nginx in foreground
4. For custom apps, ensure they don't daemonize
```
## Role Assignment
```
System: You are a senior DevOps engineer specializing in Docker security.
You are strict about security best practices and always recommend:
- Non-root users in containers
- Read-only root filesystem
- Dropping unnecessary capabilities
- Using secrets management
User: Review this Dockerfile...
```
## Structured Output
```
System: Return valid JSON only.
User: List all Docker commands for container management.
Assistant: {
"create": "docker create --name <name> <image>",
"start": "docker start <name>",
"stop": "docker stop <name>",
"remove": "docker rm <name>",
"list": "docker ps -a"
}
```
## Self-Correction
```
System: After giving your answer, evaluate it for errors.
If you find an error, say "CORRECTION:" followed by the fix.
```
## Persona Prompting
```
System: You are a Docker expert who has been using containers for 10 years.
You've seen every pitfall. Be direct and warn about common mistakes.
```
## Rule-Based
```
System: Rules:
1. Always include the exact command, not a description
2. Always include a pitfalls section
3. Never suggest docker-compose (v1), use docker compose (v2)
4. If the command needs admin, note it
```
## Pitfalls
- System prompt is most influential — invest time here
- Few-shot examples must be representative, not edge cases
- Chain-of-thought adds tokens — use only for reasoning tasks
- Structured output can fail with complex schemas
- Long prompts use context window — manage token budget carefully

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!