skills.rest
MomoDaviluke
MomoDavilukeCOMMUNITY
Explore all skills by this author →
GitHub

kb-retriever

Locate and answer questions from local Markdown, PDF, and Excel knowledge bases using layered indices and bounded retrieval.

2|Updated Mar 3, 2026
Search Tags:#knowledge-retrieval#local-kb#pdf-reading#layered-index#progressive-retrieval#excel-reading
One-click install
npx skills add https://github.com/MomoDaviluke/star-citizen-promotion --skill kb-retriever-momodaviluke
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: kb-retriever
Source: https://github.com/MomoDaviluke/star-citizen-promotion/tree/main/.agents/skills/kb-retriever
Command: npx skills add https://github.com/MomoDaviluke/star-citizen-promotion --skill kb-retriever-momodaviluke

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pdf2image, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This skill helps AI agents query a local, multi-format knowledge base efficiently by using layered data_structure.md indexes, enforcing a learn-before-process step for PDFs/Excel, and performing bounded, progressive retrieval to avoid loading entire documents into context.

Core Features & Use Cases

  • Layered index navigation: each directory can contain a data_structure.md that forms an index tree guiding the agent to relevant files.
  • Learn-before-process: when encountering PDFs or Excel, the agent must read references/*.md to learn the proper tools and parameters before processing.
  • Progressive retrieval: reads only matching windows (offset/limit) and greps to minimize token usage.
  • Bounded iterations: up to five retrieval rounds to refine results and determine when enough evidence is found.
  • Knowledge-base location flexibility: can use the default knowledge/ root or a user-specified path.

Quick Start

Place your knowledge base under knowledge/ and ask questions; the agent will navigate the layered data_structure.md, read references for PDF/Excel processing, and retrieve results in bounded rounds.

Frequently Asked Questions about kb-retriever

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I retrieve specific information from a local knowledge base without loading entire PDFs into context?▼

Progressive retrieval uses layered indices and reads only matching windows via offset and limit to extract local knowledge base answers, minimizing token usage while preserving context across bounded rounds.

What is the best way to query Markdown and Excel files in a local knowledge base?▼

Querying Markdown and Excel files requires navigating layered data_structure.md indexes and reading references/*.md to learn proper processing parameters before executing bounded, progressive grep and Read operations.

Does local knowledge base retrieval work with PDF sources and what are the prerequisites?▼

Local knowledge base retrieval supports PDF sources by enforcing a learn-before-process step that requires reading references/*.md for proper tool usage, and it depends on the pdf2image library for processing.

How many retrieval rounds are executed when searching a local knowledge base?▼

Local knowledge base searches execute up to five bounded retrieval rounds to progressively refine results and determine when sufficient evidence has been found before stopping.

How do I organize directories for effective local knowledge base retrieval?▼

Organize local knowledge base directories by placing a data_structure.md file in each directory to form a layered index tree, guiding the agent directly to relevant files instead of scanning blindly.

Can I specify a custom path for my local knowledge base instead of using the default root?▼

Local knowledge base retrieval supports custom paths, allowing you to specify an alternative directory location instead of relying solely on the default knowledge/ root folder for your sources.

Related Skills

Explore More Software Engineering Skills→
SaharCarmelSaharCarmel
2

candlekeep

Query local PDF and Markdown books with page-level citations.

Community
Advanced
hmpshmps

extract-learnings

Extract genuine learnings from completed sessions by identifying corrections and rework.

Community
Basic
JingyiWngJingyiWng

bmad-shard-doc

Split large Markdown documents into files by level-2 headings with an index.

Community
Intermediate
winsorllcwinsorllc

Embedding Search

Provides semantic search across documents with Tunrsity and Tunrsity-like features.

Community
Intermediate