Practical-Claude-Code-Skills-for-Masters-Students
OU-GC/Practical-Claude-Code-Skills-for-Masters-Students
Practical Claude Code Skills for Master's Students — paper-reader, science-confirmer, report-writer (Apache-2.0)
Instalação
npx convez-ai OU-GC/Practical-Claude-Code-Skills-for-Masters-Students1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
Practical Claude Code Skills for Master's Students
A small collection of Claude Code skills for master's students, built for academic and research work. Each skill encodes a working pattern — not a rigid checklist — so Claude reasons from the underlying principle when a new situation comes up.
Skills
paper-reader
Plain, direct explanations of research papers that keep analytical depth while dropping academic stiffness.
- Prose by default — coherent sentences for explanation; bullets only when the content is genuinely a list (parallel items, steps, specs, comparisons).
- Established terms only — uses terminology the field already recognizes; never coins term-like phrases, and says so plainly when no standard term exists.
- Conclusion first, reasoning after — one idea per sentence; concrete example before the abstract name.
- Plain voice — drop hedging qualifiers, no forced metaphors, short paragraphs, a senior-labmate tone rather than a lecturer's.
- Strictly evidence-based — grounded in what the paper actually says; name the uncertainty instead of guessing or filling gaps.
- Never drops load-bearing facts — training regime (full fine-tune / LoRA / frozen), base model and size, and whether large-scale pretraining was used.
- Fixed answer structure — one-line verdict, the key idea, why it matters, and honest limits.
paper-reader-for-chinese-user
A Chinese-user variant of paper-reader with the same analytical principles,
tuned for Chinese-speaking readers.
- Same core as
paper-reader— conclusion first, plain voice, strictly evidence-based, never drops load-bearing facts, same fixed answer structure. - Chinese gloss — English academic terms are followed by their Chinese
translation in parentheses, e.g.
attention mechanism(注意力機制). - No em-dash — uses periods, commas, colons, or parentheses instead of the
——em-dash.
report-writer
A working pattern for writing, rewriting, and polishing academic reports — Word
(.docx) documents and presentation decks (.pptx).
- Content before format — draft each section in Markdown and review it
before pouring it into the
.docx/.pptx; plan and confirm the structure before writing. - One source of truth — every term, stage name, number, and claim aligns to the canonical spec/diagram/data; fixes are applied everywhere, not just where pointed.
- Faithful to the data — write results strictly from the real numbers, no overclaiming, no jarring metaphors; surface contradictions instead of hiding them.
- Citations are claims — verify each reference actually supports what it's cited for, renumbered by order of first appearance.
- Tables & figures — numbered in order, grouped, cleaned up.
- Diagrams — architecture / flow / framework figures authored in HTML and exported to PNG, with the HTML kept as the editable master.
- Format hygiene — no emoji, no awkward justification gaps, deliberate use of space; slides designed as slides, not walls of prose.
- Manual edits are sacred — the user edits the same file by hand, so never clobber their wording.
report-writer-for-chinese-user
A Chinese-user variant of report-writer with the same writing pattern, adding
a bilingual workflow for English/Chinese reports.
- Same core as
report-writer— content before format, one source of truth, faithful to the data, citations as claims, tables/figures and format hygiene, diagrams authored in HTML, manual edits are sacred. - English first, then translate into idiomatic, academically natural Chinese — re-expression, not word-for-word mapping; the Chinese keeps the English layout identical.
- Language-correct tables — no Chinese text left inside the English version.
science-confirmer
Enforces evidence-based answers for any science, social-science, or finance claim.
- Search before you speak — find evidence before answering, preferring high-quality sources (peer-reviewed > institutional > reputable news).
- Honest about uncertainty — calibrate the wording to the strength of the evidence; lay out both sides when findings conflict.
- No fabrication — never invent technical terms, never present a hypothesis as established fact, never fill gaps with plausible-sounding guesses.
- Sound reasoning — don't confuse correlation with causation; don't overuse parenthetical asides.
Installing
Claude Code discovers skills under ~/.claude/skills/ (personal) or
.claude/skills/ (per-project). To install all of them personally:
# clone, then copy the skills into your personal skills directory
git clone https://github.com/OU-GC/Practical-Claude-Code-Skills-for-Masters-Students.git
cp -r Practical-Claude-Code-Skills-for-Masters-Students/skills/* ~/.claude/skills/
Or install just one:
cp -r Practical-Claude-Code-Skills-for-Masters-Students/skills/report-writer ~/.claude/skills/
Each skill is a directory containing a SKILL.md with YAML frontmatter
(name, description). Claude Code auto-loads it and triggers it based on the
description. Restart Claude Code (or start a new session) after copying.
Repository layout
Practical-Claude-Code-Skills-for-Masters-Students/
├── LICENSE
├── NOTICE
├── README.md
└── skills/
├── paper-reader/SKILL.md
├── paper-reader-for-chinese-user/SKILL.md
├── report-writer/SKILL.md
├── report-writer-for-chinese-user/SKILL.md
└── science-confirmer/SKILL.md
License
Licensed under the Apache License, Version 2.0. See LICENSE and NOTICE.