paper-lifecycle

Why we included this project

Most of the work in getting a paper accepted happens after the reviews come back, and this repo turns that phase into something an AI agent can run. It ships two Codex skills. The first, review-revision, reads a manuscript the way a reviewer would and flags whether the problem is real, the novelty holds, the method actually serves the mechanism, and the experiments back the claims, then hands back a section-by-section revision plan and a final checklist. The second, rebuttal-response, takes the reviews you received and sorts them into an evidence package an area chair can act on: which concerns could flip the decision, which reviewers might be persuadable, what new evidence to run, plus an AC-facing summary and a tone check that catches over-apologizing or hardball replies. It installs as a Codex skill, and Claude Code or other agents can read the skill directories directly, so it fits teams already doing agent-assisted writing. If you're prepping an A-conference or journal submission, it's a low-effort way to get a consistent second pass before you hit submit.