사용 방법
- 설치 명령을 복사하고 자신의 에이전트 환경에 설치하기 전에 소스 패키지를 확인하세요.
- 필요한 도구, 인증 정보, 지원 환경은 원본 문서에서 확인하세요.
- 입력과 원하는 산출물을 설명하고 출처와 결과를 대조하세요.
npx skills add https://github.com/vasanthsreeram/alvarmethod --skill probe시작 전 확인
- 설치에는 모델 구독, API 한도, 파일 접근 권한이 포함되지 않습니다.
- 별점 수와 설치 수는 활동 지표이며 독립적인 품질·보안 검증이 아닙니다.
원본 저장소의 제목, 예시, 코드, 표, 링크, 이미지를 유지해 표시합니다.
Probe
Measure first. Do not teach during a probe except a one-line correction after they answer.
Quiz UI is mandatory. Read ../teach/references/quiz-ui.md and call that tool. Never print A/B/C/D in the chat.
Goal
Produce .alvar/maps/<slug>.md that labels every dependency strand:
known | edge | unknown | blocked
blocked = they said they don't know, or the question needs a tool you don't have.
How to ask
- 1–3 multiple-choice questions per turn. Wait.
- Always include D. I don't know (or an extra "I don't know" option).
- Start wide (what kind of object is this? what does this operator measure?). Then split the strand that the answer leaves ambiguous.
- If they already listed solid ground, skip those strands.
- Invite a talk-through: a wrong reason + right letter is
edge, notknown.
One right answer. No trick options that hinge on wording. Put I don't know as an option. Do not mark the correct choice recommended.
Scoring
| result | status move |
|---|---|
| correct + sound reason | known |
| correct, thin reason | edge |
| wrong, near-miss | edge |
| wrong, foundation missing | unknown |
| I don't know | blocked or unknown |
After each batch, update the map. Stop when the upcoming lesson's dependencies are all labeled — not when you have asked "enough questions."
Map file
# Map — <goal>
Updated: <ISO date>
Goal: <one sentence>
## Strands
| strand | status | evidence |
|--------|--------|----------|
| … | known | Q2 + reason |
## Quiz log
- Q1 [strand] <letter> — correct|wrong|idk — <five words>End
Show the table. Say which strands a teacher should start from. Do not start the lesson unless they ask.
