What is paperjury?

PaperJury is a tool for researchers who want to stress-test their academic paper before submitting it to a conference or journal. Rather than applying every AI suggestion automatically, it sorts feedback into three outcomes: issues that can be fixed safely at the text level, issues that require the author to make a research decision (like running a missing experiment), and issues that are invalid because the reviewer misread the paper. This triage step is the central design choice. The review process is structured like a courtroom. Multiple simulated domain reviewers read the full paper, contested issues go through a two-sided trial, and a jury deliberates before returning a verdict. A verdict can conclude that no fix is needed, which a typical rewriting tool cannot produce. After fixes are applied, the paper is reviewed again from scratch in the next round. This loop continues until a clean round finds nothing new. The system tracks what changed between rounds rather than just pushing the text forward. For changes the tool deems safe, it applies them only after the author approves. Riskier edits are queued for a human review pass rather than applied silently. If you have a LaTeX toolchain installed, PaperJury can run a real compile and report actual errors, undefined references, overfull lines, and page count. Without a LaTeX installation, it falls back to structural checks and reports that limitation honestly. PaperJury installs as a skill for Claude Code, either via a marketplace command or by cloning the repo into the Claude skills folder. Once installed, you interact with it through natural language: describing a specific edit triggers the direct-edit mode, asking for a review triggers the courtroom engine, and opting in explicitly with a goal command starts an unattended review-and-revise loop. The README emphasizes that the tool is for pre-submission self-checking only and should not be used to fabricate experiments or hide limitations.