Cangjie Skill
A meta-skill that distills methodologies from books, long videos, podcasts and courses into a set of ready-to-invoke agent skills
- 12,779 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Runs the local Python tool scripts/cangjie.py and reads source content files
- Downloads packages from the repository releases, checksum verification is required
Install
- Works in
Manual install
dsh plugin --profile web add ~/.dsh/packages/dsh-cangjie-skill-2.5.0.tgz
dsh webPlugin for DeepSeek Harness. First download the dsh-cangjie-skill package and its .sha256 file from the releases page, verify the checksum, then install from the local archive.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Runs the local Python tool scripts/cangjie.py and reads source content files
- Downloads packages from the repository releases, checksum verification is required
This is third-party code. Review the repository files before installing.
What it does
Cangjie takes long-form material: a book, a video or podcast transcript, a course, an interview, and turns the methodologies inside it into atomic reusable skills. It runs the seven-stage RIA-TV++ pipeline: whole-content analysis by Adler's method, parallel extraction by five extractors (frameworks, principles, cases, counter-examples, glossary), triple verification with a promotion gate for standalone skills, capability card construction, Zettelkasten linking, pressure testing with bait questions and deterministic compilation. The output is not one summary but a skill repository with an overview, an index, a glossary, SKILL.md files and test-prompt sets. There are two build modes: a single router skill or a compact pack with a router plus promoted standalone skills. The local scripts/cangjie.py tool handles diagnostics, compilation, incremental updates, rollback and evaluation. It requires the source text: without it, distillation from memory is refused.
Who it is for. For people who read and watch a lot and want to turn methodologies from content into reusable agent skills rather than another set of notes.
Good fit when
- You want to turn a book's methodologies into a set of invokable skills, not a summary
- You have a transcript of a long video, podcast or course to build skills from
- You want a structured skill repository with trigger tests rather than one note
Not a fit when
- You just want a summary, notes or a book review
- You have no source text of the content: distillation from memory is not supported
- You want to imitate the author's style and persona rather than the methodology: that is a different tool
Example request
Distill this book into a set of executable agent skills, here is the file pathLimitations
The skill needs the source text: a book in PDF, EPUB or TXT, or a video or podcast transcript; it does not work from memory. For video and podcasts the transcript must be obtained beforehand with a separate tool. The local scripts/cangjie.py toolchain is Python and runs on your machine. Extraction quality depends on the agent's model and the completeness of the source. The refreshed package ships through the repository releases, and the archive has an sha256 checksum to verify.
How to disable. Remove the cangjie-skill/ directory from the agent's skills folder. If you installed the DeepSeek Harness plugin, remove it via dsh plugin.
FAQ
How is this different from a summary?
The goal is structured reuse, not compression. The output is several skill modules with triggers, boundaries and links, not one generalized note.
Can it distill videos and podcasts?
Yes, if you have a transcript or subtitles. Get the text first with a separate tool, then pass it to cangjie-skill.
Discussion
No questions or reviews yet.
Related
SkillsA meta-skill that takes a vague idea to a finished agent skill: it designs, writes the files, checks them and refines by evidence
A meta-skill that turns a workflow into an agent skill: prior-art search, trigger evaluation and publication to GitHub
CLI turns functions and commands from source code into verified skill records with provenance and retrieval metadata.
A meta-skill that creates, evaluates, packages and governs reusable agent skills
A skill for creating skills that proves them by behavior: before-and-after runs, regression evals and ecosystem audit
A skill that designs, tests, evaluates and packages other agent skills, starting from architecture
README in short
The README describes Cangjie as a meta-skill that distills methodologies from books, long videos, podcasts and courses into invokable skills via the RIA-TV++ pipeline. In version 2.5.0 extraction first builds a stable Capability Bundle, from which either a single router skill or a compact pack is deterministically compiled. A unified local tool, scripts/cangjie.py, covers diagnostics, compilation, updates, rollback and evaluation. The skill installs by extracting a release archive into the skills folder, with a separate plugin package for DeepSeek Harness. It supports Claude Code and OpenClaw, the README is in English, Chinese and Japanese, and it is MIT licensed.
README badge
Original SKILL.md
--- name: cangjie-skill description: 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). metadata: cangjie.version: "2.5.0" --- # cangjie-skill — 把一本书蒸馏成一组可执行 skills 的元 skill ## 使命 把一本书里沉淀的方法论,拆解成**原子化、可被 agent 在真实场景下调用**的能力,并按用户目的编译成合适数量的 skill,让读者真正用起来。 > **术语约定**: 本文档及 `methodology/`、`extractors/` 中所有的"书",泛指一切被蒸馏的长内容 — 书籍、长视频转写、播客文字稿、课程、访谈、长文、资料集。 **边界**: - ✅ 做: 方法论 / 决策框架 / 操作流程 / 计算规则 / 排障 / 清单 / 原则的蒸馏，概念体系作为参考支撑 - ❌ 不做: 书摘 / 读后感 / 作者人设角色扮演 (后者请用 nuwa-skill)