Audit a paper, codebase, benchmark, or artifact for reproducibility gaps in ML, systems, and hardware research. Use when checking whether claims can be reproduced, comparing paper-versus-code behavior, validating release readiness, or preparing an artifact evaluation package.
설치
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Audit a paper, codebase, benchmark, or artifact for reproducibility gaps in ML, systems, and hardware research. Use when checking whether claims can be reproduced, comparing paper-versus-code behavior, validating release readiness, or preparing an artifact evaluation package.
Research Reproducibility Audit
Use this skill when a result looks plausible but the path to reproducing it is weak.
Core Workflow
Enumerate the paper or artifact claims.
Map each claim to code, config, data, hardware, and evaluation evidence.
Separate what is released from what is merely described.
Identify blockers:
missing checkpoints,
hidden preprocessing,
absent seeds,
undocumented hardware,
unstable dependencies,
benchmark ambiguity.
Rank reproduction risk by impact and fix cost.
Execution Rules
Distinguish claim failure from release incompleteness.
Prefer exact paths, commands, and config names over summaries.
Treat hardware and software environment drift as first-class risks.
End with the minimum changes required for a third party to reproduce the result.
