How does it prevent duplicate notes?

It uses vector search and keyword grep to identify existing insights, then evaluates whether to skip or create an enrichment task for the new content.

Can I customize the extraction categories?

Yes, Reduce uses a derivation manifest file to map domain-specific vocabulary, extraction categories, and platform-specific hints.

What commands trigger the Reduce skill?

You can trigger the skill using the /reduce command, as well as natural language prompts like 'extract insights' or 'mine this file'.

How does it handle large files?

The skill includes a source size check and automatically plans processing in manageable chunks of 350-1200 lines to maintain context and accuracy.

What makes Reduce different from a standard summarizer?

Reduce focuses on 'articulated reasoning'—the 'why' behind information—rather than just 'what' is there, creating atomic, retrievable notes instead of generic summaries.

Reduce: Knowledge Extraction & Insights

Name: Reduce: Knowledge Extraction & Insights
Author: LopeWale

byLopeWale

•

Learning & Documentation

Extracts structured, atomic insights and articulated reasoning from source materials to build comprehensive, retrievable knowledge bases.

Reduce is a high-fidelity extraction engine for Claude Code that transforms raw source material into a structured graph of atomic propositions. Unlike traditional summarizers, Reduce focuses on 'articulated reasoning'—capturing the underlying logic and 'why' behind implementations rather than just the 'what.' It follows a comprehensive extraction principle with a strict skip rate of less than 10% for domain-relevant material, ensuring that patterns, tensions, trade-offs, and validations are never lost. By utilizing vector search for duplicate detection and enrichment, it allows developers and researchers to build robust, externalized reasoning vaults that serve as a source of truth for complex projects.

Key Features

01Vector-search enabled duplicate detection and enrichment

021 GitHub stars

03Automated large-source chunking and processing

04Comprehensive high-fidelity extraction with <10% skip rate

05Domain-specific vocabulary and configuration mapping

06Atomic insight classification (patterns, tensions, anti-patterns)

Use Cases

01Mining architectural reasoning and design patterns from legacy codebases

02Synthesizing research papers into atomic, retrievable propositions

03Building a structured knowledge vault from technical documentation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lopewale/dev_ignition reduce

For use in Claude.ai and ChatGPT

Download Skill