varunmoka7/layman
layman
「I'm lost」 또는 「explain again」 같은 문구를 말하면 Claude의 답변을 일상적인 말로 다시 표현하는 Claude Code 플러그인입니다. 로컬 hook으로 요청 횟수를 세고 지속되는 쉬운 말 모드를 켜며, 쉬운 언어 규칙을 모델 컨텍스트에 붙입니다. hook이 없는 앱에는 같은 규칙을 skill로도 제공합니다.
이 mod 소개
layman은 layman, I'm lost, explain again 같은 200자 이하의 짧은 프롬프트를 지켜봅니다. 일치하면 hook이 요청 수를 늘리고, 한 세션에서 두 번째로 요청했을 때 쉬운 말 모드를 켜서 이후 모든 답변을 일상적인 단어로 유지합니다. hook은 모델 컨텍스트에 짧은 메모를 붙입니다. 첫 줄부터 쉬운 단어를 쓰고, 한 가지 생각과 실제 예 하나, 평범한 비교 하나를 사용하며, 용어는 같은 문장에서 정의하거나 생략하고, 새 주제를 추가하지 않는다는 규칙입니다. /layman on, /layman off, /layman 명령으로 직접 제어할 수 있고, 활성화되면 상태 줄에 plain mode가 표시됩니다. 모든 기능은 로컬에서 실행되며 네트워크 호출은 없습니다. 여섯 개의 규칙 문장은 hook이 없는 Claude 앱과 Cowork에서 사용할 수 있는 skill(skills/layman/SKILL.md)로도 제공됩니다. Claude Code 2.1 이상이 필요하며 MIT 라이선스입니다.
설치
먼저 작성자의 README에서 marketplace와 플러그인 이름을 확인하세요. 저장소 구조에 따라 명령어가 달라질 수 있습니다.
claude plugin marketplace add varunmoka7/layman claude plugin install layman
원문 / README
Claude Code writes its replies for programmers. When you are not one, or it is late, a reply can be four paragraphs of words you do not know, and the easiest thing is to nod and move on.
layman is the friend who leans over and says "in other words". Type layman,
I'm lost or explain again, and the next reply comes back in everyday
words: one idea, one example from your own work, one comparison to something
ordinary, and at most one yes-or-no question. Ask twice in a session and
every reply stays plain until you switch it off.
The icon is a philosopher drawn with one line. That is the whole idea.
Install
/plugin marketplace add varunmoka7/layman
/plugin install layman@layman
Needs Claude Code 2.1 or later. Nothing else to set up.
Use
Type one of these as your prompt. Capital letters do not matter.
| Say | Also works |
|-----|------------|
| layman | lay man |
| I'm lost | im lost, i am lost |
| explain again | explain it again, explain that again |
| didn't understand | dont understand, do not understand, and the usual typos |
Only prompts of 200 characters or less count. A long bug report that quotes "I'm lost" is about something else.
| Command | What happens |
|---------|--------------|
| /layman on | Every reply stays plain. The status line shows plain mode. |
| /layman off | Back to normal. |
| /layman | Flips it. |
Plain mode also switches itself on at your second ask in a session.
Before and after
You asked why your build fails and got a paragraph about peer dependency
resolution. You type layman. The reply now reads:
Your project asked for two different versions of the same library and npm refused to pick one. It is like two people booking the same seat. Run
npm install react@18so both sides agree on one version. Want me to run it? I would.
How it works
One hook runs on every prompt you submit. It does three things:
- Count. If the prompt matches a phrase above, the ask count goes up by one.
- Decide. Two or more asks, and plain mode is on.
- Attach a note. For a matching ask, or any prompt while plain mode is on, it adds a short note to the context the model reads: everyday words from the first line, one idea, one real example, one everyday comparison, define any term in the same sentence or leave it out, no new topics.
Otherwise it adds nothing and your prompt goes through untouched. The note
itself is six sentences in
skills/layman/SKILL.md. Paste them into any
Claude chat and you have layman without installing anything.
In the Claude apps and Cowork there are no hooks, so the same six sentences ship as a skill. The model applies them when you ask for a plain explanation. There is no counter there.
Privacy
Everything runs on your computer. The plugin reads your prompt to check for a phrase and keeps one number, the ask count, for the session. No network calls, nothing stored, nothing sent.
Development
claude plugin test . # phrases match, near misses do not, plain mode switches
claude plugin validate . # plugin files are well formed
.claude-plugin/ plugin.json, marketplace.json, icon
hooks/ the hook and the /layman command
skills/layman/ SKILL.md, the rules as a skill
assets/logo/ the logo as SVG
tests/ the test suite
Both commands run on GitHub for every change. Tested on macOS with Claude Code 2.1.288.
Issues and pull requests are welcome. If you add a phrase, add it to the
hits list in tests/layman.test.ts and to the table above. Keep the plugin
small: one hook, one command, no network. Security issues go through
private vulnerability reporting.
