Skip to main content

explain(agentic-dev-kit):按问题选择清晰的讲解形式

讲解代码、流程、错误或概念时,从够用的清晰文字开始;复杂主题可用图表或 HTML 参考页。只有明确要求视频时才进入视频形式。

来源信息

仓库
ARYANK-08/agentic-dev-kit
最近来源活动
2026年10月2日 02:08
检测到的 SKILL.md 语言
英语
星标
41
分支
3

相关案例

源文档的 API client 改写示例把抽象长句改成三个具体行为:请求附带登录 token、收到 401 时退出登录、合并相同 GET 请求。它是作者的措辞示例,不代表测试过某个 API 实现。

用途

说“like I'm 13”等请求时,简化模式用同一个日常类比串起说明,再逐步解释术语和小例子。代码讲解要求先读实际仓库文件,并给出文件与行号依据。

必要前提

README 面向 Claude Code,安装后要求重启。丰富图表、HTML 页面和视频路径分别引用 Artifact 工具,以及独立的 artifact-diagramming、artifact-design、faceless-explainer Skill。

操作说明

提出讲解代码、流程、错误、概念或架构的请求,来源说明这些请求会匹配 Skill。输出先用一句话回答,再解释机制和具体输入/输出例子;需要日常类比时可说“like I'm 13”。

限制

文档采用放宽的 ASD-STE100 写作规则,没有给出正式符合该标准的认证。小问题默认用文字,视频须明确请求;丰富形式依赖文档另行引用的工具与 Skill。

安装方式

默认使用会先检查来源的 Prompt;你也可以切换为直接命令,或下载本地副本。

检查来源文件

决定是否安装前,请先阅读 SKILL.md,以及 SkillsMP 当前展示的配套文件。

正在显示 SKILL.md

SKILL.md
来源说明 · 只读预览
name
explain
description
Use whenever the user asks to explain, walk through, or help them understand anything (code, a flow, a bug, a concept, an architecture, a model output). Picks the clearest output format and writes in simplified technical English.
# Explain Goal: the user understands fast. Pick the lightest format that works, then escalate only when it helps. ## Format ladder 1. **Text** — default. Write in ~80% ASD-STE100 (see rules below). 2. **Diagram** — when the answer is a flow, structure, sequence, or relationship. Follow the Diagram style below. In the terminal use plain ASCII boxes and arrows; for a richer one, publish via the Artifact tool (`artifact-diagramming` skill). 3. **HTML page** — a one-page reference sheet (see below) when the topic is large, has many parts, or benefits from interaction (tabs, step-through, annotated code). Publish with the Artifact tool (`artifact-design` skill). 4. **Explainer video** — only if the user asks. Use the `faceless-explainer` skill. Short question → text. "How does X flow / fit together" → diagram. "Explain the whole X" → HTML page. Do not escalate to a bigger format unasked for a small question. ## Reference sheet (HTML explainer style) For "explain the whole X" or any spec, overview, or system with many parts. One page, lettered panels, like an engineering drawing sheet. - **Frame:** thin outer border with grid references (columns 1-8, rows A-D) on the edges. White page, dark text. - **Panels:** bordered cards in a grid. Each has a dark letter chip (A, B, C), a bold title, and a small mono caption on the right (e.g. "annotated examples"). - **Panel types, pick what fits:** - Structure: a tree of the parts. - Anatomy: one real example with brackets and short labels under each part. - Table: rows with a status column, check (blue) for right, cross (red) for wrong. - Limits: horizontal bars with a max marker. - History: a timeline with 3-4 dots. - **Color:** one meaning each. Red = wrong or not allowed. Blue = approved, or an annotation. Grey = secondary text. Nothing else. - **Type:** mono for examples, code, captions, and numbers. Clean sans for titles and body. - **Title block** (bottom-right): title, source, owner, sheet "1 of 1". - **Content:** every panel shows a real example, never only a definition. Wrong next to right, side by side. - Build with the Artifact tool and the `artifact-design` skill. Works in light and dark mode. Must fit a phone width by stacking panels. Use the Excalidraw style below for flows and paths. Use the sheet style for specs and overviews. Do not mix them on one page. ## Diagram style (Excalidraw look) Minimal, hand-drawn, calm. Looks like a whiteboard sketch, not a corporate slide. - **Shapes:** rounded rectangles, plain arrows, a few ellipses. No shadows, gradients, icons, or 3D. - **Lines:** hand-drawn feel. Slight wobble, 2px stroke, round caps. In SVG use an `feTurbulence` + `feDisplacementMap` filter (scale ~1.5); in HTML use `roughjs` from cdnjs. - **Colors:** black strokes on white (dark mode: light strokes on near-black). One accent color at most, plus soft pastel fills for grouping. - **Text:** hand-style font (`Caveat` or `Kalam` from Google Fonts), 2-4 words per label. Never a sentence in a box. - **Layout:** left to right or top to bottom, generous spacing, one arrow per relationship, no crossing lines. Max ~8 boxes; split the diagram if you need more. - **Arrows:** label only when the verb is not obvious ("sends token", "401"). - **Output:** inline SVG in an Artifact for rich diagrams. For editable output, write an `.excalidraw` JSON file the user can open at excalidraw.com. Terminal fallback: ``` [Browser] --token--> [app-studio] --> [optics] --> [Device] ``` ## Writing rules (ASD-STE100, relaxed) Limits: | Item | Max | | ---- | --- | | Procedural sentence | 20 words | | Descriptive sentence | 25 words | | Paragraph | 6 sentences, one topic | | Noun cluster | 3 words | | Instructions per sentence | 1 (except simultaneous actions) | Verbs: | Form | Example | OK | | ---- | ------- | -- | | Command | Close the valve. | yes | | Simple present / past / future | The valve closes. | yes | | Infinitive | Turn the knob to close it. | yes | | Past participle as adjective | The closed valve | yes | | Progressive (-ing) | The valve is closing. | no | | Perfect | The valve has closed. | no | | Passive in procedures | The valve must be closed. | no | Words: - Same word for the same thing, every time. No synonyms for variety. - Keep "the", "a", "this". Do not drop articles. - Short common words: use not utilize, start not commence, before not prior to, about not approximately, fill not replenish, make sure not ensure, to not in order to. - Active voice. Simple present tense. Use vertical lists for complex text. Safety: put the command first, then the reason. `WARNING` = risk of injury. `CAUTION` = risk of damage. Example: "WARNING: Do not touch the brake unit until it is cool. Hot parts can cause injury." Rewrite check (same meaning, 12 words instead of 16): - Before: "It is imperative that the operator ensures the hydraulic reservoir is replenished prior to commencing operation." - After: "Make sure the hydraulic reservoir is full before you start the operation." ## Structure 1. One-line answer first. 2. Then the mechanism: what calls what, with `file:line` references for code. 3. A concrete example (real input → real output) before any abstraction. 4. End with what to look at next, only if useful. Ground code explanations in the actual repo. Read the files first. Do not explain from memory. ## Simple mode (explain like I'm 13) Use when the user says "like I'm 13", "ELI13", "dumb it down", "I don't get it", or similar. This replaces the technical register; the format ladder still applies. - Start with one everyday analogy (a restaurant, a post office, a school). Keep the same analogy to the end. - No jargon. If a technical word is needed, say it once, then explain it in plain words. - Short sentences. One idea each. - Build up: what it is, why it exists, how it works, one small example. - Never talk down to the user. Simple means clear, not childish. - End with a one-line recap. ## Examples ### Bad vs good (technical mode) Question: "What does the API client do?" - Bad: "The client leverages an interceptor pattern to facilitate the approximately seamless attachment of authentication credentials prior to dispatching requests." - Good: "The client adds your login token to every request. If the server says 401, it signs you out. It also merges identical GET requests into one." Why: one idea per sentence, active voice, plain words, no stacked nouns. ### Simple mode Question: "Explain Redux like I'm 13." > Think of a school office. Every student (component) needs to know today's schedule. Instead of each student keeping their own copy, the office keeps one board (the store). A student who wants a change fills in a form (an action). The office clerk (the reducer) updates the board. Everyone looks at the same board, so nobody has old info. > > Recap: one shared board, changed only through forms. ### Pick the format | Question | Format | Why | | -------- | ------ | --- | | "Why does this hook re-render?" | Text | One cause, one fix | | "How does a request reach the device?" | Diagram: `Browser → app-studio → optics → device` | It is a path with steps | | "Explain the whole reports feature" | HTML reference sheet | Many parts: list, summary, drill-down | | "Make a video on how login works" | Explainer video | You asked for one | ### Answer shape ``` Answer: Login sends you to the SSO portal, which returns a token in the URL. How: ProtectedRoute checks the token (src/containers/ProtectedRoute.tsx:12) → AuthContext stores it → client.js sends it with each request. Example: open /modules while signed out → redirect to SSO → sign in → back at /modules with ?token=… → URL is cleaned. Next: read src/utils/ssoHandoff.js for the trust check. ```
在 GitHub 查看