medsci-skills

왜 쓰나요

핵심 기능

23 skills spanning design → analysis → manuscript → submission
33 reporting-guideline compliance audits (STROBE, PRISMA, STARD, CONSORT…)
Python/R code generation for 13+ common study designs
Verified citations — anti-hallucination by default
Built by an actively publishing physician

라이브 데모

실제 사용 모습

medsci-skill.replay ▶ 준비됨

0/0

설치

작성자 저장소 · Claude Skills 공식 문서

사용 사례

실전 활용법: medsci-skills

Draft a STROBE-compliant observational study manuscript

👤 Clinical researchers with a cleaned dataset and ready to write ⏱ ~120 min advanced

언제 쓸까: You have analyses done and need to turn them into a submittable IMRAD manuscript.

흐름

Draft methods and results from outputs

Use medsci-skills. Here are my analysis outputs and study design (retrospective cohort). Draft the Methods and Results sections, STROBE-compliant.✓ 복사됨

→ Full drafts with STROBE item mapping
Run compliance audit

Now audit this manuscript against the STROBE checklist — flag every missing item.✓ 복사됨

→ Table: item / present? / note
Self-review from reviewer perspective

Now critique this manuscript as a reviewer at the target journal would. Be harsh.✓ 복사됨

→ Numbered critique with line-level references

결과: A manuscript that clears a reporting-guideline audit and has been stress-tested before submission.

함정

Citations that look right but don't exist — The skill enforces verification — never accept a citation you can't open in PubMed

함께 쓰기: arxiv · biomcp

Run a PRISMA-compliant systematic review pipeline

👤 Researchers starting a systematic review / meta-analysis ⏱ ~240 min advanced

언제 쓸까: You're beginning a PRISMA-guided review.

흐름

Build search strategy and fetch

Use medsci-skills to design a PRISMA search for '<topic>'. Run it against PubMed and Europe PMC.✓ 복사됨

→ Boolean strategy + hit counts per database
Screen and extract

For the included studies, extract into the meta-analysis pipeline format.✓ 복사됨

→ Structured extraction table
Generate PRISMA flow diagram and forest plot

Now generate the PRISMA diagram and forest plot figures at journal resolution.✓ 복사됨

→ PDF/PNG figures ready to paste

결과: A compliant SR manuscript scaffold with figures.

함정

Over-inclusive search returns 10k+ hits — Tighten terms with the skill's query-refinement helper

함께 쓰기: arxiv

Calculate sample size with an IRB-ready justification paragraph

👤 PIs preparing an ethics submission ⏱ ~15 min intermediate

언제 쓸까: Before IRB review, when you need the stats justification done right.

흐름

Specify effect size, alpha, power

Use medsci-skills sample-size. Design: two-arm RCT. Primary outcome: continuous, expected difference 0.3 SD, alpha 0.05, power 0.8, 1:1 allocation.✓ 복사됨

→ N per arm + the formula + an IRB-ready paragraph

결과: A justification block you can paste into the protocol.

Draft point-by-point responses to reviewer comments

👤 Authors preparing a revision ⏱ ~45 min intermediate

언제 쓸까: You got R1 reviews back and need a disciplined response.

흐름

Structure the response letter

Here are the reviewer comments. Use medsci-skills to draft a point-by-point response — quote comment, response, revised manuscript location.✓ 복사됨

→ Structured table-like response with line refs

결과: A cover letter editors will thank you for.

함정

Agreeing with every comment and weakening the paper — Ask the skill to push back on comments you disagree with, with evidence

조합

다른 MCP와 조합해 10배 효율

medsci-skill + arxiv + biomcp

Pull primary literature with verifiable IDs for citations

Use biomcp + medsci-skills literature-search to find RCTs on <topic> from 2020–2025.✓ 복사됨

medsci-skill + filesystem

Keep protocol / manuscript / figures / stats in one versioned folder

Create project structure per medsci-skills conventions and initialize git.✓ 복사됨

도구

이 MCP가 노출하는 것

도구	입력	언제 호출
literature-search	topic, databases	Start of any research task
sample-size	design, effect size, alpha, power	Protocol writing
deidentify	clinical dataset	Before any data work
stats-code	design, data structure	Analysis phase
manuscript-imrad	outputs, design	Writing phase
compliance-audit	manuscript, guideline (STROBE/PRISMA/…)	Pre-submission
journal-recommender	manuscript	Picking where to submit
reviewer-response	reviews, manuscript	Revision phase

비용 및 제한

운영 비용

API 쿼터: None from the skill itself
호출당 토큰: Large — manuscript drafts run 10–30k tokens
금액: Free, OSS
팁: Draft with a smaller model, polish with Opus — the 2-pass approach halves spend

보안

권한, 시크릿, 파급범위

자격 증명 저장: No credentials stored; any DB API keys (PubMed, etc.) read from env

데이터 외부 송신: Only to databases you configure — PubMed, arXiv, etc. Never send PHI to external LLMs without de-identification.

Always run deidentify before feeding patient data to the manuscript/stats skills.
Citations must be spot-checked — the skill verifies, but humans sign the paper.

문제 해결

자주 발생하는 오류와 해결

literature-search returns no results

Broaden terms; the skill defaults to strict Boolean

Compliance audit flags many items as missing

That's expected on a first draft — work through them in order of severity

대안

medsci-skills 다른 것과 비교

대안	언제 쓰나	단점/장점
biomcp	You mainly need biomedical data retrieval without the manuscript layer	Less writing support
arxiv	Pre-prints only	No medical guideline compliance