メニュー
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.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
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.
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.
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.
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.
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.
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.
| name | proposal |
| description | 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. |
Develop the research idea in the arguments into a proposal under research/proposals/, built from templates/project-proposal.md. This is the Phase 3/4 workflow from milestones/roadmap.md.
research/reading-lists/ and paper notes (run /lit-review first if the area is unmapped). The delta must name specific papers and state what they don't do — "nobody has tried X" requires evidence of looking.Hand the draft to the skeptical-reviewer agent with instructions to attack: leakage paths, weak baselines, benchmark overfitting, p-hacking surface area, unfalsifiable claims, and scope realism. The user must revise or rebut every major point — unresolved attacks go in the proposal's own limitations section.
research/proposals/YYYY-MM-short-name.md with a timeline of weekly milestones and named kill criteria ("abandon if X hasn't worked by week N")./advisor review of the project for 2 weeks out.