Skip to content
Get packs

Cangjie Skill

Turns books, videos, and podcasts into callable AI Agent Skills.

GitHub Download v2.5.0
Stars
12k
Forks
1.4k
Licence
MIT
Language
Python
Last update
1 months ago
Type
Dev tool

What it is

Cangjie Skill takes long-form content such as books, transcripts of videos, and podcasts, and extracts the methodologies in them into structured capability cards. It then compiles these into installable Agent Skills that an AI agent can invoke when solving a task. The stated goal is to make learned material usable in real decisions rather than left in notes or watch-later lists.

  • Distill a book into a set of executable Agent Skills by passing its file path to an agent
  • Convert a long video's subtitles or transcript into a methodology skill pack
  • Compile extracted capabilities into a single router-style Skill or a multi-skill pack
  • Use the local toolchain to diagnose, update, roll back, or evaluate generated skills

Great for

  • Developers building reusable Agent Skills from a book or course they want an AI agent to apply
  • People who work with Claude Code or OpenClaw and want to package expert methods as skills
  • Readers who have text or subtitled video transcripts of a methodology-heavy source
  • Teams experimenting with knowledge distillation into agent workflows

Not ideal for

  • People who only want a summary of a book, since the project is built to produce executable skills rather than summaries
  • Users without Python 3.10+ or comfort with command-line tools, because the scripts require them
  • Anyone expecting automatic video downloading; the README recommends a separate video-downloader skill for that step
Step-by-step guide: how to use Cangjie SkillInstall, first run, examples, tips and FAQs
Cangjie Skill vs 女娲.skill

How to get it

  1. 1

    Clone and build from source

    You'll need: Python 3.10+ · PyYAML

    Install PyYAML dependency

    python3 -m pip install pyyaml

    Run the self-check

    python3 scripts/cangjie.py doctor

    Clone the repository

    git clone https://github.com/kangarooking/cangjie-skill
  2. 2

    Run it

    • Install Python 3.10+ and PyYAML.
    • Run python3 scripts/cangjie.py doctor to confirm the setup works.
    • In a new agent task, ask it to use cangjie-skill to distill a source file, giving the file path.
    • Review the generated skill output before relying on it.
  3. 3

    Get the most out of it

    • Convert video content to text first (subtitles or transcript) using the recommended video-downloader skill
    • Use the single or pack delivery modes to choose between one router Skill and multiple standalone Skills
    • Use the incremental update, diff, and rollback features when the source material changes
    • Run the evaluation and benchmarking commands in the unified toolchain to check output quality

Downloads

Source code (.zip)main branch · straight from GitHubRelease v2.5.04 files incl. cangjie-skill-2.5.0.zip.sha256

Strengths

  • Clear pipeline with defined stages, including verification and a promotion gate before output
  • Deterministic local scripts with a self-check command (doctor) and rollback support
  • Supports multiple platforms (OpenClaw, Claude Code, DeepSeek Harness) and provides checksums for releases

Trade-offs

  • Output quality depends on source text quality; the README says it works best with text, subtitles, or transcripts
  • The README is truncated, so full setup details and limits are not visible here
  • Some guidance is tied to specific release packages and refresh dates, which may confuse new users

How it compares

ProjectStarsTypeLevelUpdatedBest for
cangjie-skill (this)12kDev toolintermediate1 months agoDevelopers building reusable Agent Skills from a book or course they want an AI agent to apply
nuwa-skill37.2kAI model / agentbeginner12 days agoPeople who want an AI assistant to reason through a decision using a known thinker's frameworks
oh-my-claudecode40.5kDev toolintermediate4 days agoDevelopers already using Claude Code who want multi-agent workflows
promptfoo26.7kDev toolintermediate4 days agoDevelopers building apps on top of LLM APIs who want repeatable prompt tests
distilly26.3kDev toolintermediate24 days agoDevelopers who want an agent to answer in the style of a specific colleague or mentor
Archon23.8kDev toolintermediate4 days agoDevelopers who want AI coding runs to follow the same plan, test and review steps each time

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.