AgentSkillsCN

diataxis:explanation

适用于用户要求概念指南、架构概述、设计理由、权衡讨论、子系统工作原理背景、历史背景,或“为什么这样构建?”的文档。创建以理解为导向的内容,帮助建立心理模型——关注“为什么”,而非“怎么做”。不适用于直接执行任务(用操作指南)或事实查询(用参考)。

SKILL.md
--- frontmatter
name: diataxis:explanation
description: Use when the user asks for conceptual guides, architecture overviews, design rationale, trade-off discussions, background on how a subsystem works, historical context, or "why is it built this way?" documentation. Creates understanding-oriented content that builds mental models — the why, not the how. Do not use for direct task execution (use how-to) or factual lookup (use reference).

Explanation writer

Purpose

Write understanding-oriented documentation that helps readers see the why, context, trade-offs, and larger picture.

Explanation is not procedure and not reference. It is where reasoning belongs.

When to use

Use this skill when the user asks for:

  • conceptual guides
  • architecture overviews
  • design rationale
  • trade-off discussions
  • background on how a subsystem works
  • historical context
  • "why is it built this way?"

Do not use this skill for direct task execution or factual lookup.

What explanation should do

  • deepen understanding
  • connect ideas
  • provide context
  • discuss choices and alternatives
  • justify design decisions
  • help the reader build a coherent mental model

Boundaries

Explanation can include:

  • reasons
  • perspective
  • opinion with clear framing
  • alternatives and trade-offs
  • historical notes
  • comparisons and analogies

Explanation must not drift into:

  • procedural walkthroughs
  • API field-by-field reference
  • setup instructions

Structure

Use a structure like:

  1. Title framed around a topic
  2. Why this matters
  3. Core idea
  4. Context and background
  5. How the pieces relate
  6. Trade-offs / alternatives
  7. Common misconceptions
  8. Practical implications
  9. Further reading links to how-to or reference

Writing rules

  • discuss the subject, not the steps
  • make connections across the system
  • include background and history where helpful
  • surface assumptions
  • compare alternatives fairly
  • keep the page bounded around one topic

Good title patterns

  • About authentication in this library
  • Why the cache is write-through
  • Understanding the event pipeline
  • Design trade-offs in offline sync

Inputs to gather

  • topic to illuminate
  • audience maturity
  • design decisions and constraints
  • plausible alternatives
  • known misconceptions
  • related how-to and reference pages

Output contract

Produce explanation that:

  • improves the reader's mental model
  • helps them understand why, not just what
  • stays focused on one bounded topic
  • links out to action-oriented or fact-oriented docs for execution details

Anti-patterns to remove

  • numbered procedures
  • argument lists and schema dumps
  • generic motivational fluff
  • architecture without trade-offs
  • sprawling "everything about X" pages

Final self-check

Before returning, verify:

  • the page answers a why or about question
  • it connects ideas instead of listing steps
  • opinion is clearly presented as perspective, not disguised as fact
  • task execution details are linked, not embedded