Claude Skill

rag-knowledge-builder

Build and evaluate RAG corpora - ingest, chunk, embed, index, and spot-check retrieval quality. Use when creating knowledge bases for agents.

LLM Mart · 0 points · 0 0 0

Virus-scanned Reviewed automatically before listing.

Full trust report

Download navinspire-ia-navin-navin_skills_rag-knowledge-builder-e9c73a3.zip · 0 KB

navinspire-ia/navin 31 4 forks AGPL-3.0 Updated 4d ago

Part of navinspire-ia/navin — 182 skills

Install

skills CLI npx skills add https://github.com/Navinspire-ia/navin/tree/main/navin/skills/rag-knowledge-builder

Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install navinspire-ia-navin@llmmart

Git git clone https://github.com/Navinspire-ia/navin.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole navinspire-ia/navin collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

RAG Knowledge Builder

Overview

Garbage in, garbage out. Clean sources beat clever chunkers.

Workflow

Define the question types the RAG must answer.
Ingest sources (docs, web-extractor output, PDFs).
Chunk with structure awareness (headings > fixed blind windows).
Embed/index with the project’s vector store (note model + dims).
Evaluate with 10-20 gold questions; measure hit rate / faithfulness.
Fix gaps (missing docs, bad chunking) before tuning prompts.

Rules

Track provenance (source URL/path) on every chunk.
Exclude secrets and credentials from the corpus.

Files (navin)

SKILL.md 855 B

---
name: rag-knowledge-builder
description: Build and evaluate RAG corpora - ingest, chunk, embed, index, and spot-check retrieval quality. Use when creating knowledge bases for agents.
metadata: {"navin":{"emoji":"📚","category":"data"}}
---

# RAG Knowledge Builder

## Overview

Garbage in, garbage out. Clean sources beat clever chunkers.

## Workflow

1. Define the question types the RAG must answer.
2. Ingest sources (docs, `web-extractor` output, PDFs).
3. Chunk with structure awareness (headings > fixed blind windows).
4. Embed/index with the project’s vector store (note model + dims).
5. Evaluate with 10-20 gold questions; measure hit rate / faithfulness.
6. Fix gaps (missing docs, bad chunking) before tuning prompts.

## Rules

- Track provenance (source URL/path) on every chunk.
- Exclude secrets and credentials from the corpus.