Smart Content Summarizer

Problem

When engineers process meeting notes, papers, product pages, or repository docs, the main cost is not reading itself, but locating core conclusions, parameters, dates, and owners. summarize compresses source material into a structured digest: it extracts core claims, important data, and terminology, then preserves source, time, and naming cues so you can re-check the original material without rereading everything.

Capabilities and limits

The skill targets three input types: raw text, local documents, and web pages. For text, it supports material up to roughly 100,000 characters and can use short, medium, or long summaries. For documents, it handles .txt, .md, .docx, and .pdf, while preserving outline, layout clues, and important details when possible. For web pages, it filters noise such as ads and navigation, and retains metadata like source links and publication time when available. Output can be plain text, Markdown, or JSON; keywords are ranked by importance and the count can be controlled with --keywords-count. Results are saved to the current working directory, usually with file names containing keywords or timestamps for easier backup and tracking. This is best for “read conclusions first, then verify sources” workflows, not for final factual verification, contract review, or complete citation. Items marked with [!] should be cached separately according to the material’s guidance.