chain-of-summaries-cos

Generate dense, question-driven summaries with synthetic Q&A pairs from source documents.

1|Updated Apr 5, 2026

Search Tags:#summary #iterative #document #dialectical #cos #chain-of-summaries

Ready to run

npx skills add https://github.com/Alexmacapple/alex-claude-skill --skill chain-of-summaries-cos

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: chain-of-summaries-cos
Source: https://github.com/Alexmacapple/alex-claude-skill/tree/main/chain-of-summaries-cos
Command: npx skills add https://github.com/Alexmacapple/alex-claude-skill --skill chain-of-summaries-cos

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

COS provides a robust method to produce dense, gap-aware summaries by detecting lacunae through synthetic questions, revealing missing details and enabling thorough QA-ready outputs.

Core Features & Use Cases

Iterative refinement: starts with a thesis summary, then asks targeted questions to expose gaps, and refines the final text.
Question-driven coverage: generates synthetic Q&A pairs derived from the source to ensure complete factual coverage.
Versatile applicability: suitable for long documents, technical reports, and transcripts that require dense, accurate summaries.

Quick Start

Provide a source document and run the COS workflow to generate a dense, question-driven summary with iterative refinement.

Frequently Asked Questions about chain-of-summaries-cos

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

FAQPage Schema

How do I generate a dense summary that identifies gaps in long-form documents?▼

To generate a dense, gap-aware summary for long-form documents, apply an iterative Chain of Summaries (COS) workflow that detects missing details through synthetic questions and refines the text. This produces a high-density, QA-ready output.

What is the best way to summarize technical reports to ensure complete factual coverage?▼

The best way to summarize technical reports for complete factual coverage is using question-driven iterative refinement. This method generates ten synthetic Q&A pairs from the source to expose lacunae and ensure thorough, QA-ready factual outputs.

Can I use this iterative summary method on transcripts and long texts?▼

Yes, you can apply iterative summarization to long-form texts, technical reports, and transcripts. The process generates a dense, QA-ready output consisting of a thesis summary, synthetic Q&A pairs, and iterative refinements.

How does a question-driven summary workflow expose missing details?▼

A question-driven summary workflow exposes missing details by generating synthetic Q&A pairs derived from the source document. This dialectical process identifies lacunae, ensuring complete factual coverage and preventing circularity in the final dense text.

What are the limitations of using COS for document summarization?▼

Limitations of COS for document summarization include a strict adherence requirement to routing rules to avoid circularity, a targeted output length of around 200 words, and the necessity of providing a source document input to function.

Do I need to provide a source document to generate a QA-ready summary?▼

Yes, you must provide a source document as input to generate a QA-ready summary. The workflow processes this input to produce a dense, gap-aware output consisting of a thesis summary, ten synthetic Q&A pairs, and iterative refinements.