Cangjie Skill turned 300 pages of a Python book into 47 Claude Code skills

I tried a new tool that distills long-form content (books, videos, podcasts) into executable AI Agent Skills. The 3,352 GitHub stars suggest I am not the only one interested in this category. After 1 week of testing, here is what it does well and where it falls short.

The problem

I consume a lot of technical content: documentation, video courses, blog posts. The friction is: I read a 300-page book, take some notes, and then a month later I cannot remember the key patterns. Cangjie Skill addresses this by extracting the book into structured prompts that AI agents can use to perform related tasks.

How it works

Upload a PDF, paste a YouTube URL, or provide a webpage. Cangjie extracts the content, identifies the key concepts, and generates a set of skills. For a Python web development book, it generated 47 skills covering common tasks like "set up Flask with PostgreSQL" and "implement REST API with authentication".

The results I tested

I tested on 3 content types:

• 300-page Python book → 47 skills, well-calibrated to the book's style
• 8-hour YouTube Rust course → 23 skills, mixed quality
• 20-article Docker blog series → 12 skills, well-calibrated

The pattern: technical content with clear structure works best. Long-form narrative or creative content has lower quality output.

How I use it

The skills integrate with Claude Code, Cursor, or any agent that supports custom instructions. I set up a project with the extracted skills, and now I can say "set up auth in this new project" and get a response that matches the patterns from the book I read. For a developer who learns from documentation, this is the difference between re-reading 300 pages and getting a working implementation in 5 minutes.

The limitations

• Long content (>500 pages) takes significant time to process
• The distillation quality depends on the LLM backend (GPT-4 > Claude > smaller local models)
• The generated skills are static — they do not update as the source content changes
• No version control for the generated skills
• Limited language support (works best for English)

The verdict

Cangjie Skill is the most useful content distillation tool I have tested in 2026. The 3,352 stars and the open source design make this a good choice for technical self-learners. For casual content consumers, the value is less clear. But for anyone who has bought a technical book and never finished it, this tool can help.

Full review: https://saas.pet/reviews/cangjie-skill-review