How does this skill help manage the Claude context window?

It provides principles for conciseness and instruction hierarchy, ensuring critical information is prioritized while removing redundant tokens to keep conversations efficient.

Can these patterns be used for sub-agents?

Yes, these patterns are specifically designed to improve communication between a primary agent and its sub-agents or other LLM interactions.

What is the benefit of Chain-of-Thought prompting?

Chain-of-Thought forces the model to reason through a problem step-by-step before answering, which significantly reduces errors in complex logic and mathematical tasks.

What are 'Degrees of Freedom' in prompt design?

Degrees of Freedom refer to how much creative liberty you give the AI. Low freedom is for fragile tasks like database migrations, while high freedom is for tasks like creative writing or code reviews.

When should I use Few-Shot learning instead of text instructions?

Few-Shot learning is most effective when you need a specific, rigid output format or when a pattern is easier to demonstrate via examples than to describe in rules.

Prompt Engineering Patterns

Name: Prompt Engineering Patterns
Author: prof-ramos

byprof-ramos

•

開発者ツール

Optimizes LLM performance and reliability using advanced prompting techniques, structural patterns, and systematic refinement strategies.

This skill provides a comprehensive framework for designing high-performance prompts for Claude Code and other LLM-driven applications. It implements proven industry patterns such as Chain-of-Thought, Few-Shot learning, and progressive disclosure to maximize reasoning accuracy while maintaining strict control over model outputs. By focusing on instruction hierarchy and context window efficiency, this skill helps developers build more robust agents, reduce token costs, and handle complex edge cases through structured error recovery and template-based design.

主な機能

01Token-efficient system prompt and template design

02Progressive disclosure strategies to manage instruction complexity

03Few-Shot learning patterns for consistent output formatting

04Chain-of-Thought reasoning for complex multi-step logic

051 GitHub stars

06Instruction hierarchy for predictable model behavior

ユースケース

01Minimizing token usage and latency in high-volume agent interactions

02Improving LLM accuracy on analytical and mathematical reasoning tasks

03Developing production-ready prompt templates for autonomous agents

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add prof-ramos/.codex prompt-engineering

For use in Claude.ai and ChatGPT

Download Skill