skills.rest
lfuuu
lfuuuCOMMUNITY
Explore all skills by this author →
GitHub

summarize

Generates structured multi-page summaries from large DOCX and Markdown documents via a multi-phase pipeline.

Updated Jun 2, 2026
Search Tags:#summarization#pdf#markdown#fact-checking#pandoc#document analysis#docx
One-click install
npx skills add https://github.com/lfuuu/claude-rules --skill summarize-lfuuu
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: summarize
Source: https://github.com/lfuuu/claude-rules/tree/main/global-skills/summarize
Command: npx skills add https://github.com/lfuuu/claude-rules --skill summarize-lfuuu

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve? Condensing a large Word or Markdown document (project documentation, reports, specifications) into a coherent 3-4 page summary without losing key facts is slow and error-prone when done manually or in a single pass. This Skill decomposes the document into sections, builds auxiliary artifacts (map, glossary, digests, fact registry), prioritizes core content, and synthesizes a reviewed, humanized summary exported to Markdown and PDF. ## Core Features & Use Cases - Multi-phase summarization pipeline: DOCX-to-Markdown conversion (pandoc), section splitting, parallel subagent digests, fact registry, priority grading, synthesis, review-and-fix loop, humanization, and PDF export (pandoc + wkhtmltopdf). - Structure and outline modes: --structure produces a section tree with readiness status (ready/partial/in-progress/empty) and a completeness percentage; --outline adds per-section content summaries. - Document consistency checking (--check): deterministic reference checks (broken figure/table captions, dangling mentions, duplicate numbers) plus subagent-based contradiction detection with refute-verification, producing a PDF issue report. - Media handling: inventories embedded images, converts EMF/GIF to PNG, and generates vision-based descriptions of diagrams used as facts in the pipeline. - Use Case: Given a 200-page project specification in .docx, run the skill to receive a 3-4 page Russian-language summary PDF, a glossary, a fact registry, and a verified section-numbering mapping back to the original document. ## Quick Start Ask the assistant to summarize the document at a given path, for example: summarize the file 'docs/specification.docx' into a structured summary with PDF output.

Frequently Asked Questions about summarize

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

FAQPage Schema
How do I summarize a large DOCX document into a short PDF?▼

Run the summarize skill with the path to the .docx file. It converts the file to Markdown with pandoc, splits it into sections, builds digests and a fact registry via parallel subagents, synthesizes a 3-4 page summary, reviews it, and exports a PDF using pandoc and wkhtmltopdf.

How can I check a document for broken figure and table references?▼

Use the --check mode, which runs deterministic checks for broken captions, duplicate numbers, dangling mentions, and page references, then applies subagent-based contradiction detection with refute-verification. The result is a check-report in Markdown and PDF.

What tools are required to convert DOCX to Markdown and PDF?▼

The pipeline requires pandoc for DOCX-to-Markdown conversion and wkhtmltopdf for rendering the final PDF with Cyrillic support. Optional converters like inkscape, soffice, or ImageMagick handle EMF, WMF, and GIF image conversion.

Can I get a section readiness overview instead of a full summary?▼

Yes, the --structure mode builds a tree of all sections and subsections tagged as ready, partial, in-progress, or empty, with an overall completeness percentage. The --outline mode adds brief content summaries per top-level section.

How does the pipeline avoid losing context on very large documents?▼

The document is split into per-section files, and each section is processed by a separate parallel subagent that returns a structured digest, terms, and facts. Synthesis then uses only these auxiliary documents rather than the raw source, keeping the context window manageable.

What are the limitations of automated document summarization pipelines?▼

The summary is strictly limited to facts present in the source document and its auxiliary artifacts; nothing is inferred. Quality depends on correct section numbering from the original TOC, and page-level references cannot be verified in Markdown, only against the original DOCX.

Related Skills

Explore More Content & Communication Skills→
sg6724sg6724

docxmaker

Convert markdown content into Word .docx documents.

Community
Basic
crisweber2600crisweber2600

lens-reporting-snapshot

Generates read-only stakeholder status snapshots in Markdown and JSON from Lens governance artifacts.

Community
Intermediate
wise-clawwise-claw
10

source-command-subject-multilingual-extractor

Extract the primary subject from text and translate it into seven languages.

Official
Intermediate
prefactordevprefactordev
3

write-span-summary-templates

Create Liquid templates that render concise Prefactor span summaries.

Official
Intermediate