skill-from-notebook

Provide the article as a local file or URL, and the Skill first validates whether it contains a repeatable methodology. It then classifies the content as How-to, Decision, Framework, or Checklist and generates a structured skill with steps, quality checkpoints, and common pitfalls.

Provide a finished work such as a blog post or README, and the Skill analyzes its structure, proportions, and quality characteristics. It then deduces the production steps and key decisions needed to create similar output, producing a reusable template.

It supports local PDF, Word, and Markdown files read directly, web URLs fetched for content, YouTube videos via yt-dlp subtitles or Whisper transcription, and NotebookLM links through browser automation. Finished examples of any text-based artifact can also be reverse engineered.

Content lacking a clear goal, repeatable steps, quality criteria, or scenario context is unsuitable, such as news reports or pure opinion pieces. The Skill validates suitability first and honestly explains why extraction is not possible rather than forcing a low-quality result.

Yes, providing multiple examples of the same type enables multi-example learning. The Skill extracts commonalities as the core methodology and analyzes differences as style variants or optional techniques, producing more generalizable results.