Plain Speak: Jargon into Plain Language

Turn a technical term, code snippet, or dense engineering paragraph into one line your PM, exec, or customer can actually repeat back. It works from a single anchor — who the reader is — and holds every rewrite to a "could this person repeat it in a meeting?" bar.

Install

npx skills add https://github.com/leoluyi/skills -g -a plain-speak -y

Update later with:

npx skills update plain-speak

Source

What it does

Pins down the actual reader first — a CFO, a PM, a salesperson, a customer, or (by default) a non-technical manager who knows the product but not the stack — because the same term gets a different plain-language version depending on who's listening. It then produces a repeat-test line: what the thing does, and what it's for, with any unavoidable jargon glossed inline (idempotent(冪等)) rather than left to carry meaning on its own. It replies in whatever language the user wrote the request in.

It also reviews an already-written plain-language draft against the same bar, marking each criterion pass or fail and fixing what fails, instead of rewriting blind.

Invoked mid-conversation with nothing attached, the target becomes the preceding turn: it re-explains the last substantive answer, or — if a question was just put to the user and left unanswered — re-poses that question in plain language, option by option, so the user can just answer it.

When to use

Reach for it to explain a technical term, code snippet, error, or dense engineering text to a non-technical reader, to check whether a plain-language draft actually lands, or to have the answer or question you just got re-done in plain language.

When not to

Not for de-AI voice cleanup (use humanizer-zh), for structuring a whole formal document like a memo or report (use formal-doc-structure), or for an RFP (use rfp-writing).

How it works

This skill lowers the audience — it doesn't touch voice, structure, or document type. That's a deliberate boundary against three sibling skills that operate on different axes of the same prose:

humanizer-zh strips AI-sounding phrasing from already-finished text — a voice-layer cleanup, not an audience shift.
formal-doc-structure organizes a whole formal business document into the right sections for its type — a structural job, not a translation job.
rfp-writing enforces the specific conventions of an RFP document — a document-type job, distinct from lowering language for a reader.

A passage can need any one of these independently of the others: jargon aimed at a PM can still be in an AI-ish voice, or correctly plain but structurally disorganized. Plain Speak only handles the audience axis.

Related skills

humanizer-zh — owns removing AI-isms and polishing tone in prose; this skill doesn't touch voice.
formal-doc-structure — owns organizing a whole formal document like a memo or report; this skill only lowers language, not structure.
rfp-writing — owns RFP-specific conventions; those requirements conflict with a plain-language rewrite aimed at a general reader.

把術語、程式碼片段或工程長文，翻成非技術的主管、業務或客戶聽了能直接複述的一句話。它先鎖定一個錨點——讀者是誰——再拿「這個人能不能在會議上複述」當唯一的品質標準。

安裝

npx skills add https://github.com/leoluyi/skills -g -a plain-speak -y

之後更新：

npx skills update plain-speak

原始碼

它做什麼

先把讀者鎖定成具體的人——CFO、PM、業務、客戶，或預設情況下懂產品但不懂技術堆疊的非技術主管——因為同一個術語，講給不同的人聽，白話版本會不一樣。接著產出一句能複述的話：這東西做什麼、是為了解決什麼問題，任何避不開的術語都就地標註（例如「idempotent（冪等）」），而不是讓術語自己承載語意。回覆一律用使用者提問時所用的語言。

它也能審查一份已經寫好的白話草稿，逐條標記通過或不通過，再針對不通過的地方修改——而不是不看標準就整段重寫。

在對話中間空手呼叫、沒附任何內容時，目標就是上一輪對話：重講上一個有實質內容的回答；如果上一輪是丟給使用者、還沒被回答的問題，就把那個問題逐個選項用白話重新問一次，讓使用者可以直接回答。

何時使用

要把技術術語、程式碼片段、錯誤訊息或工程長文講給非技術讀者聽，想檢查白話草稿夠不夠白，或想叫它把剛剛那個回答、那個問題用白話重講一次時使用。

何時不要

不要用於潤飾語氣、去 AI 味（用 humanizer-zh），編排整份簽呈或報告這類正式文件（用 formal-doc-structure），或 RFP（用 rfp-writing）。

運作方式

這個技能降低的是讀者門檻——不動語氣、不動結構、也不動文件類型。這是刻意畫出的邊界，對照三個處理同一段文字不同軸線的兄弟技能：

humanizer-zh 處理已經寫完的文字裡的 AI 腔調——是語氣層的清理，不是換讀者。
formal-doc-structure 把整份正式商業文件組織成該類型該有的章節——是結構的工作，不是翻譯的工作。
rfp-writing 執行 RFP 文件特有的規範——是文件類型的工作，跟為讀者降白話是兩回事。

同一段文字可能同時需要這三者中的任何一個，彼此獨立：給 PM 看的術語說明，講法可以已經很白話、卻還是 AI 腔；也可能語言已經夠白，結構卻很亂。Plain Speak 只處理讀者這一個軸。

Plain Speak: Jargon into Plain Language把技術術語翻成白話