What problem does it solve? When a finished paper is rejected at one venue, moving it to another venue usually means risky manual edits: reformatting to a new LaTeX style, absorbing prior reviewer concerns, shrinking page count, and preserving anonymity — all without accidentally mutating the original submission or introducing new experiments, citations, or framework changes. This Skill automates that entire constrained resubmission workflow. ## Core Features & Use Cases - Physical isolation setup: Composes the new venue submission as a sibling directory via physical copies, guaranteeing prior submission directories are never overwritten. - 5-layer anonymity scan: Detects surface identifiers, self-citation phrasing, acknowledgments, cross-rebuttal references, and internal codenames before double-blind submission. - Whitelist-gated microedits: Runs an auto-improvement loop bound by an edit whitelist that forbids bib edits, new theorems, new citations, and unbacked numerical claims, with per-round diff checkpoints. - Adversarial gate and audit chain: Coordinates proof-checker, claim-audit, citation-audit (soft-only), and kill-argument skills, emitting a forensic RESUBMIT_REPORT ledger. - Use Case: Your ICLR submission was rejected with three reviewer reports. Point the Skill at the paper directory, the NeurIPS template, and the review corpus; it produces a compiled NeurIPS-format submission with reviewer concerns addressed through text-only edits and a full audit trail. ## Quick Start Run the resubmit pipeline on my rejected paper directory targeting NeurIPS, using the folder of prior reviewer reports as the review corpus.
