kangarooking/cangjie-skill

把书、长视频、播客等高价值内容蒸馏成可执行的 Agent Skills（Distill high-value content from books, long-form videos, podcasts, and more into executable Agent Skills）

What it solves

Cangjie Skill addresses the problem of "knowledge decay" where high-value content from books, long-form videos, podcasts, and courses is consumed but never applied. It transforms passive consumption (summaries and notes) into structured, executable AI skills that an agent can invoke in real-world scenarios to solve specific problems.

How it works

The project uses a pipeline called RIA-TV++ to convert source text into a "Capability Bundle" and then into installable skills. The process consists of seven stages:

  1. Whole-Content Comprehension: Uses Adler's method for structural and critical analysis.
  2. Parallel Extraction: Five specialized extractors pull frameworks, principles, cases, and glossaries.
  3. Triple Verification: Candidates are filtered and promoted to independent entry points based on evidence and utility.
  4. RIA++ Capability Construction: Content is structured into capability cards (Reading, Interpretation, Appropriation, Execution, and Boundary).
  5. Zettelkasten Linking: Encodes dependencies and compositions in a capability graph.
  6. Pressure Testing: Uses bait questions and confusion tests to ensure the skill's reliability.
  7. Deterministic Compilation: Compiles the bundle into either a single router-style skill or a compact pack of standalone skills.

Who it’s for

Users who want to turn educational or methodological content into functional tools for AI agents, as well as developers building agentic workflows that require specialized, verified knowledge bases derived from expert literature.

Highlights

  • Execution-Oriented: Focuses on structured reuse rather than narrative compression or simple summaries.
  • Comprehensive Source Support: Works with books, video transcripts, podcasts, and interview notes.
  • Deterministic Toolchain: Includes a local toolchain for diagnostics, compilation, and incremental updates.
  • Extensive Library: Provides a wide array of pre-distilled skill packs ranging from investment reasoning (Buffett) to organizational design (Netflix).
  • Platform Integration: Offers plugins for platforms like DeepSeek Harness, Claude Code, and OpenClaw.

Related

  • Dispatch
    Combatting AI Brain Drain: Deliberate Skill Development with Claude Code

    Explore how to use AI 'skills' to transform coding assistants from passive generators into active tutors, preventing the loss of codebase knowledge and cognitive atrophy.

  • Project
    kangarooking/kangarooking-skills

    A collection of structured, open-source Agent Skills and workflows for AI agents to automate tasks ranging from content creation and 3D generation to complex software engineering.

  • Project
    yaojingang/yao-open-skills

    A collection of professional AI Skills designed to turn complex tasks like business analysis and strategic decision-making into reusable AI assets and structured reports.

  • Project
    alchaincyf/huashu-skills

    huashu‑skills is a curated index of 52 plug‑in “Agent Skills” (design, data, writing, video, persona emulation, etc.) that work with Claude Code, Codex, Kimi Code and similar AI agents. Skills are either standalone repos (install via `git clone`) or lightweight sub‑folders in this repo (copy to `~/.claude/skills/`). A machine‑readable `skills.json` lets agents discover them, and an included updater checks for newer versions.

  • Project
    Pluviobyte/rnskill

    A comprehensive library of 55 AI Agent skills for Codex and Claude Code that automates content creation, video production, and business diagnosis workflows.