tk-mwhat
Repeats the last explanation in short, clear words, but only when someone directly asks for it again.
Installation
Paste this into Claude Code, Cursor, or any agent that can run commands.
What this skill does
What it does:
This is a small helper that re-explains the assistant's last answer in a shorter, simpler way, but only when you directly ask something like 'what?' or 'what do you mean?'. It looks only at the explanation that was just given and rewrites it as a couple of very short lines, plus a short recommendation if there is one. It keeps exact wording for things like file names, commands, links, and on-screen button text exactly as they were, without translating or rephrasing them. After giving the short re-explanation, it stops completely.
When to use it:
- You did not understand the assistant's last explanation and want it said again, more simply
- You want a fast recap instead of scrolling back up to reread it
It will not edit files, run commands, make plans, or do anything besides give that one short re-explanation. It does nothing if you ask about something other than the answer given right before it, or if there is nothing recent to re-explain.
SKILL.mdShow the author's original SKILL.md (not in English)
---
name: tk-mwhat
description: "[user] 직전 설명이 이해되지 않아 사용자가 명시적으로 다시 설명을 요청할 때만, 바로 앞 explanation/source를 짧고 정확하게 재설명한다. 자동 호출하지 않으며 파일 변경·구현·실행을 하지 않는다."
disable-model-invocation: true
argument-hint: "<직전 설명>"
metadata:
tigerkit:
kind: user-invoked
origin: tigerkit
relationship: adapted
---
# Re-explain the Previous Explanation
Use only when the user **explicitly** requests it with `/tk-mwhat` or phrases such as `뭐라고?`, `뭐라는 거야?`, or `무슨 말이죠?`. The input is only the **immediately preceding explanation/source** in the conversation. If none exists or the reference is ambiguous, output exactly `Unverifiable` and stop.
## Output
Output only these two blocks.
```md
🤔 쉽게 말하면
[뜻과 현재 문제를 최대 3개의 짧은 줄로 재설명]
💡 추천
[추천 또는 그대로 보낼 다음 문장 최대 2개의 짧은 줄]
```
Write more briefly than the source without changing its meaning. Preserve paths, commands, URLs, literals,
verified UI strings, `Status`/IDs, and original authorship attribution exactly. Keep a verified UI string's language,
case, punctuation, and spacing; do not translate, paraphrase, or normalize it. Do not turn an `enum`, code identifier,
or i18n key into a UI label. If there is no basis for a recommendation, use
`추천: 없음`.
After the re-explanation, **`hard stop`**. Do not modify files, implement anything, create plans or summaries, or run commands or tools. New tasks, change requests, and general summary requests are outside this skill; stop with `NotApplicable`.
## Source
Prior precedent is TigerKit `mwhat` (commit `c6963e8`, `skills/mwhat/SKILL.md`):
two short Korean explanation blocks, exact source/literal preservation, and no implementation contract.
Matt Pocock's original `wait-what` was not available in this checkout or its records, so no behavior was inferred.
Ships with 3 supporting files:
- agents/openai.yaml
- evals/evals.json
- evals/triggers.json
Mirrored from the author's public source. Install counts from the open skills registry.