What problem does it solve? Turning scattered experiment results, review feedback, and research narratives into a coherent, venue-compliant paper outline is time-consuming and error-prone. This Skill automates that planning step by mapping claims to evidence and producing a page-budgeted outline for ICLR, NeurIPS, or ICML submissions. ## Core Features & Use Cases - Claims-Evidence Matrix: Extracts core claims from NARRATIVE_REPORT.md, GPT54_AUTO_REVIEW.md, and experiment JSON files, then maps each claim to supporting evidence and target sections. - Venue-Aware Structure: Selects empirical, theory, or method paper templates with page budgets matching ICLR (9), NeurIPS (9), or ICML (8) main-body limits. - Figure and Citation Planning: Produces a detailed figure plan (including hero figure comparisons) and a verified citation scaffolding with [VERIFY] flags. - Automated Cross-Review: Sends the outline to GPT-5.4 via Codex MCP for scored feedback on flow, claim alignment, and page feasibility before finalizing. - Use Case: After finishing an auto-review loop on your NVRC compression experiments, run this Skill to produce PAPER_PLAN.md with section breakdowns, figure plans, and reviewer feedback ready for drafting. ## Quick Start Ask the assistant to run the paper-plan skill on your project directory to generate a PAPER_PLAN.md outline from your narrative report and experiment results.
