cangjie-skill
Distill a book, long-video transcript, podcast, course, or interview into a coherent set of executable skills. Use when the user asks to "拆书" / "蒸馏一本书" / "把 XX 书做成 skill" / "把这个视频/播客/课程蒸馏成 skill" / "turn a book or video into skills" — i.e. wants the frameworks, principles, and methodologies in long-form content extracted into atomic, reusable Claude skills that an agent can invoke in real-world situations. NOT for simple summarization, book reviews, or role-playing as the author (that is nuwa-skill's job).
cangjie-skill — Distill a Book into Executable AI Skills
Skill Overview
cangjie-skill is a book-deconstruction and distillation skill that extracts methodologies from books, long-form video transcripts, podcasts, courses, and interviews, turning them into a set of atomic skills that AI agents can directly invoke in real-world scenarios.
Use Cases
- Deconstructing classic books: Break down methodology-dense books such as Poor Charlie’s Almanack into decision frameworks, checklists, calculation rules, and other executable capabilities, so you can put them into practice immediately after reading.
- Distilling audio and video content: Extract structured skill cards from transcripts of long-form Bilibili videos, podcasts, and online courses, instead of leaving behind a pile of lengthy documents that are merely stored and never read.
- Consolidating team knowledge: Convert interview records and internal training courses into a shared library of standardized skills that new team members can invoke directly.
Core Features
- Five-stage distillation pipeline: From Adler-style whole-book comprehension, through parallel extraction of candidate methodologies by five agents, triple-verification screening, and the independent Skill Advancement Gate, to final compilation and delivery. Each stage has a quality gate, supports resumable execution, and does not fabricate content from memory.
- RIA++ capability cards and unified compilation: Each methodology is structured as a complete capability card containing its intended use, inputs and outputs, examples from the original book, distinctions from adjacent capabilities, and evaluation criteria. These cards are then uniformly compiled into either a single-entry (
single) or multi-skill package (pack) delivery format. - Triple verification and stress testing: Every capability must pass three types of verification—source sufficiency, executability, and task improvement—and complete stress testing on real tasks before it can be delivered. Rejected items retain a complete audit trail, with no inflated coverage claims.
Frequently Asked Questions
How is cangjie-skill different from an ordinary AI summary?
Ordinary summaries produce “book-report-style” text that gets archived after reading. cangjie-skill produces executable skills: each methodology is broken down into an atomic capability card with explicit trigger conditions, inputs and outputs, and evaluation cases. AI agents can invoke these capabilities directly in real-world tasks instead of merely restating the book’s ideas.
What types of content are supported?
Books (PDF / EPUB / TXT), long-form video transcripts, podcast transcripts, course materials, interview records, and other long-form content. For videos and podcasts, obtaining a transcript first is recommended. If no transcript is available, it will stop and ask you for the source material rather than distilling content “from memory.” Before starting each project, it is recommended to pilot the process on one piece of content.
Can it write book reviews or imitate an author’s style?
No—this is a deliberate boundary. cangjie-skill only distills methodologies, decision frameworks, operating procedures, and principles. For book reviews and author-persona roleplay, use nuwa-skill instead. Distilled skills follow the darwin-skill evaluation format and can be connected to the automated evolution workflow with one click for continuous iteration.