Can I manage token limits with this skill?

Yes, it provides a ContextWindowManager that supports sliding window logic and smart truncation to ensure conversations stay within token limits while preserving the most relevant history.

What is Chain-of-Thought prompting in this context?

It provides a standardized pattern to force the LLM to output its reasoning steps before providing a final answer, which is proven to increase accuracy for complex logic tasks.

Does it support structured data validation?

Absolutely. It utilizes Pydantic to ensure that prompt templates are correctly filled and validated before being sent to the model.

How does this skill help with prompt injection?

It includes a dedicated sanitizer that detects dangerous instruction-override patterns and wraps user input in XML-style tags to maintain clear boundaries between system instructions and user data.

Prompting Patterns

Name: Prompting Patterns
Author: ricardoroche

byricardoroche

0•

データサイエンスとML

Implements standardized prompt engineering patterns to ensure reliable, secure, and high-performance LLM interactions.

Prompting Patterns is a specialized skill designed for developers and AI engineers building LLM-powered applications. It provides a library of implementation patterns for structured templates, few-shot learning, and Chain-of-Thought reasoning to significantly improve model output quality. The skill also addresses critical production concerns by offering robust prompt injection prevention mechanisms and sophisticated context window management strategies, ensuring your AI agents remain secure and context-aware even during complex, long-running interactions.

主な機能

010 GitHub stars

02Structured prompt templates with Pydantic validation and type safety

03Automated few-shot learning builders for consistent example formatting

04Pre-configured Chain-of-Thought patterns to improve reasoning capabilities

05Advanced prompt injection sanitization using pattern matching and XML-style wrapping

06Context window management including sliding windows and smart truncation

ユースケース

01Building robust RAG applications with complex document context requirements

02Developing secure AI agents that resist user-level instruction overrides

03Standardizing prompt design and versioning across large engineering teams

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add ricardoroche/ricardos-claude-code prompting-patterns

For use in Claude.ai and ChatGPT

Download Skill