WUBING2023

ReproRun — Development skill for Claude Code

Development community

🔬 A skill that automatically reproduces the numerical claims of academic papers — read paper → find code → build env → run → compare against the paper.

How to install ReproRun

This entry records only its repository, not the path inside it, so there is no exact command to give. Open WUBING2023/ReproRun and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What ReproRun does

🔬 A skill that automatically reproduces the numerical claims of academic papers — read paper → find code → build env → run → compare against the paper. 1 orchestrator + 6 specialized agents.

Alternatives in Development

PaperSpine — PaperSpine is a motivation-driven skill for learning from strong academic papers, building a paper’s central a 5k ★
Ars Abstract — ARS academic-paper abstract-only mode — bilingual abstract + keywords 4.5k ★
Paper Search Pro — Academic literature discovery as a Skill — Claude Code · Codex · any agent that loads SKILL.md 145 ★

README

English · 简体中文 · Español · Français · Deutsch · 日本語 · 한국어 · Português · Русский

ReproRun

Give it a paper. It tells you whether the results actually reproduce.

**ReproRun** is a portable **AI-agent skill** — usable by any compatible AI coding assistant — that automates the painful path from *"a paper claims X"* to *"X actually runs on my machine."* Papers ship beautiful numbers; reproducing them usually dies in broken code, unbuildable environments, and version drift. ReproRun handles the whole pipeline end to end:

**read paper → find code & data → build environment → smoke test → full run → compare measured numbers against the paper's claims.**

✨ Built to adapt — across every platform

ReproRun is designed to **adapt itself to whatever it's handed**, instead of assuming one fixed setup:

Cross-OS — runs on Windows, macOS, and Linux
Cross-language — full pipeline for Python; minimal path for R / MATLAB / Julia
Cross-domain — single-cell biology, image ML, and more, with no per-domain rewiring
Cross-mode — tool mode (pip install + write a calling script) or experiment mode (clone the repo & run its scripts) — auto-selected per paper

🚀 What it does

Find a paper from just its title — auto web-search for PDF & code repo
Auto-diagnose environment bit-rot — numpy ABI clashes, torchvision API changes, deprecated pandas methods… detected and fixed automatically
No guessing parameters — inspects function signatures after install
5-round dependency self-healing loop — classify error → targeted fix → re-verify, up to 5 rounds
Paper isolation — every paper gets its own output names