本文へ移動
cccskills
無料GitHub で公開

setup-medsci

Use when a skill fails for a missing tool or the environment needs checking. Diagnoses Python, R, Node, Claude Code, Git, Zotero and configured MCP servers and prints a pass/fail table with links to the setup docs. Read-only; installs nothing.

インストール方法を見る

含まれるファイル(3)

  • SKILL.md3.8 KB
  • references/setup-checklist.md3.8 KB
  • skill.yml1.1 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Setup-MedSci Skill

Diagnose only. Never install, upgrade or configure anything — no brew install, winget install, pip install, Rscript -e 'install.packages(...)', and no edits to ~/.claude.json or any MCP configuration (read claude mcp list only) — because auto-installers for system Python and R are a support nightmare for non-developer users. If a tool is missing, the user follows the setup doc themselves. Skill versions and content are out of scope (validate_skills.sh, /manage-project status).

The diagnostic table is in English so it can be pasted into a GitHub issue. Remediation links point to the docs/setup/ guides in the medsci-skills repository, and only these five exist: docs/setup/README.md, docs/setup/mac.md, docs/setup/windows.md, docs/setup/mcp-setup.md, docs/setup/common-issues.md.

Workflow

Phase 1: Detect OS

uname -s
  • Darwin → macOS → docs/setup/mac.md
  • Linux → Linux (similar tooling to Mac) → docs/setup/mac.md
  • MINGW*, MSYS*, CYGWIN*, or detection failure on Windows → docs/setup/windows.md

Phase 2: Run Diagnostic Commands

Read references/setup-checklist.md and run every check in it — R1-R5 required, O1-O5 optional — with its detect command, version command, minimum version and status rule. Use command -v <tool> first to detect presence without running it (avoids triggering long initialization). Capture the version output and the exit code, and report the version exactly as printed: if a command fails, report the failure verbatim rather than a version that is "probably" installed. If command -v finds the tool but the version command or its parse fails, mark the row ❌ and report the failing command — the install is broken, not just outdated. An MCP server is ✅ only when claude mcp list shows it connected; configured-but-disconnected is ⚠️.

Phase 3: Emit Checklist Table

Print a single Markdown table to stdout in this exact format, one row per check, with the doc link from the reference for the detected OS:

## MedSci Skills Setup Diagnostic

OS detected: <macOS | Linux | Windows>
Date: <YYYY-MM-DD>

| Component | Status | Detected | Required | Action |
|---|:---:|---|---|---|
| Python 3.11+ | ✅ / ❌ | 3.11.9 | 3.11+ | OK / See docs/setup/mac.md Step 2 |
| MCP: zotero | ✅ / ❌ / ⚠️ | Connected | optional | OK / See docs/setup/mcp-setup.md |

Summary: <X required components passed, Y missing>
Next step: <one-sentence action>

Status legend:

  • ✅ Present and meets minimum version
  • ❌ Missing or below minimum version
  • ⚠️ Present but optional and not connected

Phase 4: Suggest Remediation

Use the reference's Summary Rules for the closing sentence.

  • Everything required ✅ → point to Demo 1: cd ~/medsci-skills/demo/01_wisconsin_bc && claude '/orchestrate --e2e'.
  • Any required row ❌ → print its doc link and tell them to follow that step. Do not offer to install; the doc tells them exactly what to copy-paste.
  • Only optional MCP rows ❌ → explain that MedSci Skills work without MCP servers, but lit-sync, verify-refs and write-paper are smoother with Zotero MCP; link docs/setup/mcp-setup.md.

Output

The diagnostic report goes to stdout. If the user asks for a copyable report (e.g. for a GitHub issue), also write it to ~/.medsci-skills/diagnostic-YYYY-MM-DD.md and tell them where it is.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Use when a medical AI paper should be found and cited by AI search engines and RAG tools. Optimizes the title, abstract, summary box (Key Points, Research in Context), keywords, preprint, GitHub README/CITATION.cff and Hugging Face card, returning a visible pass/fail checklist.

日本語の概要は準備中です。原文の説明を表示しています。

Aperivue/medsci-skills3362026年10月12日 更新

Use when a target journal has no profile yet. Reads the journal's author guidelines and writes a detailed /write-paper profile plus a compact /find-journal profile, public or user-local private, in the canonical format with quality gates.

日本語の概要は準備中です。原文の説明を表示しています。

Aperivue/medsci-skills3362026年10月12日 更新

Use when data needs statistical analysis. Runs reproducible Python/R code for Table 1, diagnostic accuracy, agreement, regression, survival, propensity score, survey-weighted and repeated-measures models, with publication tables. Sample size is /calc-sample-size; pooling studies is /meta-analysis.

日本語の概要は準備中です。原文の説明を表示しています。

Aperivue/medsci-skills3362026年10月12日 更新

Renamed to /model-selection in v6 (removed in v7).

日本語の概要は準備中です。原文の説明を表示しています。

Aperivue/medsci-skills3362026年10月12日 更新

Use when analyzing a researcher's publication record from PubMed. Fetches an author's papers, classifies study types and author position, charts the patterns and writes a strategy report, with an optional trajectory-archetype classification. Works from PubMed metadata only.

日本語の概要は準備中です。原文の説明を表示しています。

Aperivue/medsci-skills3362026年10月12日 更新

Use when one validated cohort analysis must be repeated across many exposure/outcome pairs. Generates one R/Python script per combination from a single methodology template, changing only the variables, and aggregates the results into a summary matrix.

日本語の概要は準備中です。原文の説明を表示しています。

Aperivue/medsci-skills3362026年10月12日 更新

Aperivue のスキルをすべて見る

このスキルの問題を報告する