一键在 Manus 中运行任何 Skill

oral-exam

星标0

分支0

更新时间2026年6月9日 22:28

Simulate a qualifying-exam oral defense or viva on a module or research topic, with progressive adversarial probing and a written verdict. Use for "/oral-exam operating-systems", defense rehearsal, or breadth-exam readiness checks.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

fingerskier

fingerskier/CSAI_PhD

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

SKILL.md

readonly

name	oral-exam
description	Simulate a qualifying-exam oral defense or viva on a module or research topic, with progressive adversarial probing and a written verdict. Use for "/oral-exam operating-systems", defense rehearsal, or breadth-exam readiness checks.

Oral Exam Simulation

Run a qualifying-exam-style oral examination on the topic in the arguments. Default length: 15 minutes of focused questioning (roughly 6–10 exchanges). For full breadth-exam rehearsal, run one round per section of assessments/breadth-exam.md.

Use the examiner agent (.claude/agents/examiner.md) as the questioning persona, or adopt it directly in conversation.

Protocol

Open broad — one foundational question a strong student answers in 2–3 minutes.
Probe deep — follow the user's actual answer, not a script. Chase every hedge, undefined term, or hand-wave: "you said X — prove it", "what breaks if that assumption fails?", "give me the constant factors".
Force tradeoffs — present a scenario where the textbook answer fails and require the user to reason at the boundary.
Cross-connect — at least one question linking this topic to another core module (the roadmap requires demonstrated connections across ≥3 areas).
Failure modes — end by asking where the user's own understanding is weakest; examine exactly that.

Rules of engagement

One question at a time. Never answer your own question; silence or "I don't know" gets a narrower follow-up, not the answer.
Interrupt politely if an answer drifts — real committees do.
No teaching during the exam. Hold corrections for the debrief.

Verdict and debrief

After the session, deliver:

Verdict — pass / conditional / not yet, against the standard in assessments/breadth-exam.md and assessments/rubrics.md.
Transcript review — each question, what was strong, what a committee would flag.
Gap list — appended to the error log under research/, each with the module section to restudy.
Communication score — the roadmap's bar: explain any concept in ≤5 minutes with formal precision. Say whether they met it.

同仓库更多 Skills

同仓库

advisor

fingerskier/CSAI_PhD

Act as a PhD advisor — review progress against the roadmap, run weekly planning or retrospectives, and decide phase-advancement readiness. Use for "/advisor", weekly check-ins, "am I ready for Phase 2?", or when feeling stuck/unfocused.

2026-06-090

flashcards

fingerskier/CSAI_PhD

Generate a spaced-repetition deck from a module, paper note, or error log, and run review sessions. Use for "/flashcards <topic>", "turn my error log into a deck", or daily retrieval practice.

2026-06-090

lit-review

fingerskier/CSAI_PhD

Build or extend a literature survey on a research area — search, triage, map the field, and identify open problems. Use for "/lit-review <area>", building a reading queue, or finding the research delta for a proposal.

2026-06-090

fingerskier/CSAI_PhD

Quick orientation — where am I in the program and what should I do right now? Use for "/next", "what should I work on?", "where was I?", or when a new learner doesn't know where to start. For full planning meetings use /advisor instead.

2026-06-090

paper

fingerskier/CSAI_PhD

Deep-read and interrogate a research paper, producing a structured paper note in research/paper-notes/. Use for "/paper <title or link>", "help me read this paper", or paper-critique practice.

2026-06-090

proposal

fingerskier/CSAI_PhD

Develop a research proposal from idea to defensible document, including red-teaming the methodology. Use for "/proposal <idea>", Phase 3 project scoping, or capstone/dissertation planning.

2026-06-090

name	oral-exam
description	Simulate a qualifying-exam oral defense or viva on a module or research topic, with progressive adversarial probing and a written verdict. Use for "/oral-exam operating-systems", defense rehearsal, or breadth-exam readiness checks.

Oral Exam Simulation

Use the examiner agent (.claude/agents/examiner.md) as the questioning persona, or adopt it directly in conversation.

Protocol

Open broad — one foundational question a strong student answers in 2–3 minutes.
Probe deep — follow the user's actual answer, not a script. Chase every hedge, undefined term, or hand-wave: "you said X — prove it", "what breaks if that assumption fails?", "give me the constant factors".
Force tradeoffs — present a scenario where the textbook answer fails and require the user to reason at the boundary.
Cross-connect — at least one question linking this topic to another core module (the roadmap requires demonstrated connections across ≥3 areas).
Failure modes — end by asking where the user's own understanding is weakest; examine exactly that.

Rules of engagement

One question at a time. Never answer your own question; silence or "I don't know" gets a narrower follow-up, not the answer.
Interrupt politely if an answer drifts — real committees do.
No teaching during the exam. Hold corrections for the debrief.

Verdict and debrief

After the session, deliver:

Verdict — pass / conditional / not yet, against the standard in assessments/breadth-exam.md and assessments/rubrics.md.
Transcript review — each question, what was strong, what a committee would flag.
Gap list — appended to the error log under research/, each with the module section to restudy.
Communication score — the roadmap's bar: explain any concept in ≤5 minutes with formal precision. Say whether they met it.