# How to Construct a Skill Using R/I/A1/A2/E/B Structure in cangjie-skill

> Learn to build AI skills with cangjie-skill using the R I A1 A2 E B structure. Understand each component for effective AI development.

- Repository: [kangarooking/cangjie-skill](https://github.com/kangarooking/cangjie-skill)
- Tags: how-to-guide
- Published: 2026-08-15

---

**The cangjie-skill framework builds executable AI skills from book methods using a six-part RIA++ structure: Reading (R), Interpretation (I), Past Application (A1), Future Trigger (A2), Execution (E), and Boundary (B).**

The **cangjie-skill** repository transforms distilled method-units from books into **Claude Code skills** through a rigorous six-section model called **RIA++**. Each section serves a distinct pedagogical and functional purpose, ensuring the resulting skill is both grounded in source material and immediately actionable by AI agents. This architecture is defined in the methodology specification at [[`methodology/04-stage2-ria-plus.md`](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md)](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md) and rendered through [`templates/SKILL.md.template`](https://github.com/kangarooking/cangjie-skill/blob/main/templates/SKILL.md.template).

## The Six Components of R/I/A1/A2/E/B Structure

### R — Reading: Source Extraction

The **Reading** section provides a verbatim excerpt from the original source, strictly limited to ≤150 characters or ≤100 English words. It must include precise citation (chapter, page, or timestamp) and remain "raw"—never pre-translated—to satisfy copyright-friendly usage rules.

According to [`methodology/04-stage2-ria-plus.md#L11-L16`](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md#L11-L16), this anchoring step ensures every skill traces back to a specific location in the source book.

```markdown

## R — 原文 (Reading)

> "当你想要做出正确的决策时，先假设相反的结果会怎样。"  
> — 查理·芒格, 第三讲

```

### I — Interpretation: Own-Words Explanation

The **Interpretation** section rewrites the core idea in the author's own words, spanning 5-15 lines. Direct copying from the source is prohibited—this section must enable a naïve reader to grasp the method without consulting the original book.

As specified in [`methodology/04-stage2-ria-plus.md#L17-L22`](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md#L17-L22), this step tests true comprehension rather than mechanical transcription.

```markdown

## I — 方法论骨架 (Interpretation)

逆向思维是一种先假设最坏结果的思考方式，帮助我们识别潜在风险并反向校正决策。
通过把注意力从"我该怎么做"转向"如果我不这样做会怎样"，可以逼出隐藏的假设，
防止盲目乐观。

```

### A1 — Past Application: Author's Historical Usage

**A1** documents one to three concrete examples where the book's author actually applied the method. Each example follows a four-part structure:

- **Problem** — The situation requiring the method
- **Method's usage** — How the method was deployed
- **Conclusion** — The insight gained
- **Concrete result** — The measurable outcome

The specification at [`methodology/04-stage2-ria-plus.md#L24-L30`](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md#L24-L30) requires these examples to be drawn directly from the source text, not invented.

### A2 — Future Trigger: Activation Logic

The **A2** section is the **most critical** component for AI execution. It defines:

1. **When** users need the skill (3-5 vivid scenarios)
2. **Language signals** — phrases users utter that should fire the skill
3. **Initial distinction** from neighboring skills (refined in stage 3)

Per [`methodology/04-stage2-ria-plus.md#L32-L41`](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md#L32-L41), this content is collapsed into the front-matter `description` field with a ≤300 word limit, making it the primary activation signal for Claude agents.

```yaml
---
name: reverse-thinking
description: |
  用户在纠结一个决策、列举正面理由却理不出头绪时;或在问"怎么做 X 才能成功"时。
  不适用于纯信息查询或日常琐碎选择。
---

```

### E — Execution: Step-by-Step Recipe

The **Execution** section delivers a concrete 1-2-3 step recipe. Each step requires:

- **Action description** — What to do
- **Measurable completion criterion** — How to know it's done
- **Optional stop-gate logic** — Conditional jumps (e.g., "if X then jump to step 5")

The specification at [`methodology/04-stage2-ria-plus.md#L53-L58`](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md#L53-L58) emphasizes that these steps must be executable without ambiguity.

```markdown

## E — 可执行步骤 (Execution)

1. **列出所有可能的正面方案**  
   - 完成标准: 至少三条明确的正向方案。
2. **逆向假设每条方案的最坏结果**  
   - 完成标准: 为每条方案写出对应的失败情景。  
   - 判停条件: 若所有最坏结果均不可接受，则直接放弃该方案。
3. **比较正反两面，挑选风险最可控的方案**  
   - 完成标准: 选出风险收益比最高的方案。

```

### B — Boundary: Usage Constraints

The **Boundary** section explicitly defines where the skill **must not** be used. It draws from:

- The book's own warnings about misuse
- Stage-0 critique identifying limitations
- Known overlaps with other methods to prevent over-calling

Per [`methodology/04-stage2-ria-plus.md#L61-L68`](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md#L61-L68), this protective layer prevents harmful advice and reduces false activations.

```markdown

## B — 边界 (Boundary)

### 不要在以下情况使用此 skill

- 仅需事实查询（如"今天的天气如何？"）。
- 已经有明确的法律或安全约束（如"是否可以非法下载？"）。

### 作者在书中警告的失败模式

- 只做逆向思考而忽视正向验证，导致过度保守。

### 容易混淆的邻近方法论

- 与 "假设检验" 的区别: 本 skill 侧重个人决策情境，而假设检验为统计学工具。

```

## Skill Construction Pipeline

The R/I/A1/A2/E/B structure feeds into an automated generation process:

1. **Input**: Verified method-units from `books/<slug>/verified.md`
2. **Template**: [`templates/SKILL.md.template`](https://github.com/kangarooking/cangjie-skill/blob/main/templates/SKILL.md.template) defines required front-matter and markdown layout
3. **Output**: [`SKILL.md`](https://github.com/kangarooking/cangjie-skill/blob/main/SKILL.md) written to `books/<slug>/<skill-slug>/`

The resulting skill file is **self-contained and executable**—loadable by Claude agents, testable in stage 4, and linkable to other skills in stage 3.

## Summary

- **R/I/A1/A2/E/B** is the mandatory six-section structure for all cangjie-skill skills, defined in [[`methodology/04-stage2-ria-plus.md`](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md)](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md).
- **R** anchors to source text with strict length limits; **I** demands original explanation.
- **A1** proves historical efficacy; **A2** drives AI activation through front-matter `description`.
- **E** provides measurable steps; **B** prevents misuse through explicit constraints.
- The [**`templates/SKILL.md.template`**](https://github.com/kangarooking/cangjie-skill/blob/main/templates/SKILL.md.template) renders verified units into executable Claude Code skills.

## Frequently Asked Questions

### What does RIA++ stand for in cangjie-skill?

**RIA++ expands to Reading, Interpretation, A1 (Past Application), A2 (Future Trigger), E (Execution), and B (Boundary).** The "++" distinguishes the six-part model from simpler three-part RIA frameworks. According to the cangjie-skill methodology, A2 and E/B were added specifically for AI agent activation and safe execution.

### How is the A2 section used by Claude agents?

The **A2 (Future Trigger)** content populates the front-matter `description` field, which serves as the primary signal for when Claude should invoke the skill. It contains scenario descriptions, trigger phrases, and skill distinctions—all compressed into ≤300 words for rapid matching against user queries.

### Why does the Boundary section matter for R/I/A1/A2/E/B skills?

The **B (Boundary)** section prevents harmful over-application by explicitly listing forbidden use cases, author-warned failure modes, and overlaps with neighboring skills. Per [`methodology/04-stage2-ria-plus.md#L61-L68`](https://github.com/kangarooking/cangjie-skill/blob/main/methodology/04-stage2-ria-plus.md#L61-L68), this protects users from advice that might be technically correct but contextually inappropriate.

### Can I create a cangjie-skill without all six sections?

**No.** The [`templates/SKILL.md.template`](https://github.com/kangarooking/cangjie-skill/blob/main/templates/SKILL.md.template) enforces all six sections as mandatory. Skills missing R, I, A1, A2, E, or B will fail the stage-4 testing pipeline and cannot be linked in stage 3.