How does it handle token limits?

The skill includes logic for context window management that fits the most relevant documents into a specified token budget while reserving room for the system prompt.

What is a RAG sufficiency check?

A sufficiency check is a pre-generation step that evaluates if the retrieved context contains enough information to answer the query, which prevents the LLM from hallucinating when data is missing.

Does this skill help with source citations?

Yes, it provides specific implementation patterns for tracking document metadata and generating responses with inline citations like [1] and [2].

How does hybrid search work in this skill?

It uses Reciprocal Rank Fusion (RRF) to combine the results of semantic (vector) search and keyword search, ensuring high precision and recall.

RAG Retrieval Patterns

Name: RAG Retrieval Patterns
Author: yonatangross

byyonatangross

•

Data Science & ML

Implements production-ready Retrieval-Augmented Generation patterns to ground AI responses in factual data and minimize hallucinations.

The RAG Retrieval skill provides a comprehensive toolkit for building high-performance, grounded AI applications. It offers standardized patterns for basic retrieval, inline citation tracking, and hybrid search techniques like Reciprocal Rank Fusion (RRF). Beyond simple search, it introduces advanced 2026 best practices for context sufficiency checks, which evaluate retrieved data before generation to prevent the model from guessing when information is missing. This skill is essential for developers building documentation bots, knowledge management systems, or any AI tool requiring verifiable factual accuracy.

Key Features

01Reciprocal Rank Fusion (RRF) for optimized search ranking

02Intelligent context window and token budget management

0369 GitHub stars

04Pre-generation sufficiency checks to eliminate hallucinations

05Automated citation tracking and source attribution logic

06Hybrid search patterns combining semantic and keyword results

Use Cases

01Developing a search-to-answer pipeline for large-scale internal data repositories

02Building an enterprise knowledge base chatbot with verifiable sources

03Creating a documentation assistant that refuses to answer when information is absent

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yonatangross/orchestkit rag-retrieval

For use in Claude.ai and ChatGPT

Download Skill