Ailog RAG Skills for Claude Code

Professional skills for building, auditing, evaluating, and optimizing RAG (Retrieval-Augmented Generation) systems with Claude Code.

Overview

These skills help you build production-grade RAG pipelines by providing:

Quick Start

Installation

Option 1: Via Claude Code Marketplace (Recommended)

# Add the marketplace
/plugin marketplace add https://github.com/floflo777/claude-rag-skills

# Install all skills
/plugin install rag-audit
/plugin install rag-eval
/plugin install chunking-advisor
/plugin install rag-scaffold

Option 2: Manual Installation

# Clone the repository
git clone https://github.com/floflo777/claude-rag-skills.git

# Copy to your Claude Code skills directory
cp -r claude-rag-skills/* ~/.claude/skills/

# Or for project-specific installation
cp -r claude-rag-skills/* .claude/skills/

Usage

After installation, use the skills in any Claude Code session:

You: /rag-audit
Claude: I'll analyze your codebase for RAG-related code and check for anti-patterns...

You: /chunking-advisor
Claude: What types of documents will you be indexing? What embedding model are you using?

You: /rag-scaffold
Claude: I'll help you generate a production-ready RAG pipeline. What's your preferred framework?

You: /rag-eval
Claude: Let's evaluate your RAG system. Do you have a test dataset, or should I help create one?

Skills Documentation

/rag-audit - RAG Code Auditor

Scans your codebase for RAG implementations and identifies:

• Chunking issues: Wrong size, no overlap, boundary problems
• Embedding problems: Model mismatch, no caching, batch issues
• Retrieval anti-patterns: Fixed top-k, no reranking, missing hybrid search
• Generation issues: Context overflow, poor prompts, no citations
• Production gaps: Missing error handling, logging, caching

Example output:

# RAG Audit Report

## Summary
- Files Analyzed: 12
- Issues Found: 8 (2 critical, 4 warnings, 2 suggestions)
- Overall Score: 72/100

## Critical Issues

### No Chunk Overlap
**Location**: `src/chunker.py:45`
**Issue**: Chunks created with overlap=0
**Impact**: Information at chunk boundaries will be lost
**Fix**: Add 10-20% overlap

/rag-eval - RAG Evaluator

Evaluates your RAG system with standard metrics:

Retrieval Metrics:

• Recall@K, Precision@K
• Mean Reciprocal Rank (MRR)
• Normalized Discounted Cumulative Gain (NDCG)

Generation Metrics:

• Faithfulness (grounded in context)
• Relevance (answers the question)
• Coherence and conciseness

Optional: Ailog Benchmark

Compare your system against Ailog's production RAG API:

export AILOG_API_KEY="pk_live_your_key"
export AILOG_WORKSPACE_ID="123"

/chunking-advisor - Chunking Strategy Expert

Get recommendations based on:

• Document type (code, legal, FAQ, articles, tables)
• Query patterns (factual, analytical, comparative)
• Embedding model (token limits, optimal sizes)
• Performance requirements

Decision tree included for quick strategy selection.

/rag-scaffold - RAG Boilerplate Generator

Generate complete, production-ready RAG pipelines:

Framework Options:

• Python + LangChain + Qdrant
• Python + LlamaIndex
• Python Vanilla (no framework)
• TypeScript + LangChain.js
• Ailog API (managed RAG)

Includes:

• Configuration management
• Embedding service with caching
• Vector store operations
• Retrieval with reranking
• Generation with streaming
• Docker setup
• Tests

Ailog Integration

These skills reference Ailog's RAG guides for best practices:

• Chunking Strategies
• Choosing Embedding Models
• Hybrid Search
• Reranking
• RAG Evaluation
• Production Deployment

Optional API Integration:

The /rag-eval skill can benchmark against Ailog's API for objective comparison. Create a free workspace at ailog.fr to use this feature.

Project Structure

claude-rag-skills/
├── rag-audit/
│   └── SKILL.md           # Audit skill instructions
├── rag-eval/
│   └── SKILL.md           # Evaluation skill instructions
├── chunking-advisor/
│   └── SKILL.md           # Chunking advice instructions
├── rag-scaffold/
│   └── SKILL.md           # Scaffold generation instructions
├── examples/
│   └── ...                # Example configurations
├── marketplace.json       # Plugin marketplace metadata
└── README.md

Requirements

• Claude Code >= 2.0.0
• For Ailog benchmarking: Ailog API key (optional)

Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Submit a pull request

For major changes, open an issue first to discuss.

License

MIT License - see LICENSE for details.

Support

• Documentation: Ailog Guides
• Issues: GitHub Issues
• Discord: Ailog Community

Built with expertise from Ailog - The RAG-as-a-Service Platform