What Do the R, I, A1, A2, E, B Fields in SKILL.md Signify? A Complete Field Guide

The R, I, A1, A2, E, B fields in SKILL.md represent a structured six-part documentation template—Reading, Interpretation, Past Application, Future Trigger, Execution, and Boundary—used to codify skills from source texts for AI agent consumption.

The cangjie-skill repository, maintained by kangarooking, enforces this rigid field structure to transform book-derived wisdom into machine-actionable capabilities. Every skill must declare all six sections, verified through automated quality checks, before downstream agents like darwin-skill can invoke it.

---

The RIA-TV++ Framework: Six Mandatory Fields Explained

The field names abbreviate a methodology documented in methodology/04-stage2-ria-plus.md. Here's what each field requires:

R — Reading (原文)

Purpose: Anchor the skill in verifiable source material.

This field contains a direct quotation from the original text, capped at 150 words or 150 Chinese characters, with proper attribution. It grounds the skill in primary evidence rather than paraphrase or memory.

In templates/SKILL.md.template, this section appears as ## R — 原文 (Reading) and must include blockquote formatting with citation.

I — Interpretation (方法论骨架)

Purpose: Translate the author's framework into clear, self-contained explanation.

Here, the skill author rewrites the underlying method in 5–15 lines, using their own words. This bridges the gap between original context and actionable abstraction.

The pipeline enforces that I cannot merely repeat R—it must demonstrate independent synthesis.

A1 — Past Application (书中的应用)

Purpose: Establish credibility through documented case studies.

This field catalogs concrete examples from the source material where the author applied the method. Each case typically includes:

• The problem encountered
• How the methodology was deployed
• The conclusion reached
• The result or outcome

A1 proves the method has been battle-tested, not just theorized.

A2 — Future Trigger (触发场景) ★

Purpose: Define activation conditions for AI agents.

This critical field specifies exactly when an agent should invoke the skill. It includes:

1. Situational descriptions (e.g., "user hesitating over a high-price purchase")
2. Explicit language signals—verbatim phrases that trigger recognition (e.g., "I'm hesitating," "what are the risks")

The ★ marker in the template indicates this field is essential for agent routing logic.

E — Execution (可执行步骤)

Purpose: Convert abstract principles into concrete actions.

E provides 1–3 step-by-step instructions with success criteria. Unlike I, which explains what the method is, E dictates how to perform it.

Steps must be imperative, measurable, and agent-executable.

B — Boundary (边界) ★

Purpose: Prevent misuse through explicit scope limitation.

The final field declares where the skill fails:

• Inapplicable situations (e.g., "not for pure information queries like weather checks")
• Author-documented failure modes from counter-examples
• Contextual blind-spots the agent must recognize

The ★ marker emphasizes that boundary definition is mandatory for safe deployment.

---

SKILL.md Structure in Practice

Below is a complete, runnable template excerpt showing all six fields in their required format, derived from templates/SKILL.md.template:

---
name: effective-decision-making
description: |
  When users face important decisions, need to list positive reasons but get stuck; or ask "how to do X to succeed". Not for daily trivial choices.
source_book: 《穷查理宝典》 Charlie Munger
source_chapter: Chapter 12
tags: [decision-making, mental-models]
related_skills: []
---

# Effective Decision Model

## R — Original Text (Reading)

> "When facing complex decisions, first write down all positive reasons, then invert each reason for thinking." — Charlie Munger, Chapter 12

---

## I — Methodology Framework (Interpretation)

Break decisions into two steps: ① List all positive reasons; ② Invert-verify each reason to uncover hidden risks...

---

## A1 — Application in Book (Past Application)

### Case 1: Investment Decision

- **Problem**: When to buy...
- **Methodology use**: ...  
- **Conclusion**: ...  
- **Result**: ...  

---

## A2 — Trigger Scenarios (Future Trigger) ★

1. User hesitating over expensive purchase.
2. User asks "how to evaluate risk".

**Language signals**  
- "I'm hesitating"  
- "what are the risks"

---

## E — Executable Steps (Execution)

1. **List positive reasons** – ...  
2. **Invert thinking** – ...  

---

## B — Boundary (Boundary) ★

- Not for pure information queries like "today's weather".
- Author's warning failure mode: ignoring negative factors leads to blind optimism.

---

Quality Enforcement: How Fields Are Validated

The pipeline enforces R/I/A1/A2/E/B completeness through "质量红线" (quality red lines) defined in methodology/04-stage2-ria-plus.md. Key validation rules:

Validation Check	Implementation
Field presence	All six headings must exist with exact ## X — format

| Content non-emptiness | No field may contain only whitespace or placeholder text | | A2 trigger format | Must contain numbered scenarios and bolded "Language signals" subsection | | B scope denial | Must explicitly state 不适用于 or equivalent negation | | Citation in R | Source must be attributed with —— or comma-delimited author/location |

Files missing any field fail verification and cannot be indexed by darwin-skill or other downstream consumers.

---

Key Files in the cangjie-skill Repository

Understanding these source locations clarifies how the field system operates:

• templates/SKILL.md.template — Canonical structure definition with all six fields pre-templated
• methodology/04-stage2-ria-plus.md — Stage 2 pipeline documentation describing RIA-TV++ population and validation
• methodology/00-overview.md — Methodology overview contextualizing the full extraction pipeline
• SKILL.md — Root documentation referencing quality red lines

These files collectively ensure that "skill" in cangjie-skill means a validated, semantically-structured, agent-ready capability—not merely a note or summary.

---

Summary

• R (Reading) — Direct, cited quotation from source (≤150 words)
• I (Interpretation) — Author's 5–15 line framework explanation
• A1 (Past Application) — Historical case studies proving method validity
• A2 (Future Trigger) — Activation conditions with verbatim language signals
• E (Execution) — 1–3 actionable steps with success criteria
• B (Boundary) — Explicit scope limitations and failure mode warnings

All six fields are mandatory, validated by automated checks in methodology/04-stage2-ria-plus.md, and structured according to templates/SKILL.md.template.

---

Frequently Asked Questions

What happens if a SKILL.md file is missing one of the six fields?

The pipeline rejects it. The "质量红线" (quality red lines) in methodology/04-stage2-ria-plus.md enforce complete R/I/A1/A2/E/B presence. Files failing this check cannot be consumed by darwin-skill or other downstream agents.

Why are A2 and B marked with ★ in the template?

These fields are operationally critical: A2 determines when agents activate the skill, and B prevents harmful misapplication. The ★ symbol in templates/SKILL.md.template flags them for extra scrutiny during manual review and automated validation.

Can the field order be changed or reordered in SKILL.md?

No. The template in templates/SKILL.md.template prescribes fixed sequence R → I → A1 → A2 → E → B. Pipeline parsers expect this ordering; deviations may cause validation failures or incorrect indexing.

How long should each field typically be?

Field	Target Length
R	≤150 words/Chinese characters
I	5–15 lines
A1	1–3 case studies
A2	2–5 scenarios + language signals
E	1–3 steps
B	2–5 scope limitations

These constraints ensure skills remain concise enough for agent context windows while retaining actionable detail.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}