What problem does it solve? Turning raw research artifacts—review conclusions, experiment logs, and narrative reports—into a coherent, venue-compliant paper outline is slow and error-prone. This Skill automates that planning step, producing a claims-evidence-backed outline with page budgets, figure plans, and citation scaffolding. ## Core Features & Use Cases - Claims-Evidence Matrix: Extracts core claims from NARRATIVE_REPORT.md, AUTO_REVIEW.md, and experiment results, mapping every claim to supporting evidence. - Venue-Aware Structure: Supports ICLR, NeurIPS, ICML, CVPR, ACL, AAAI, and IEEE venues with correct page-limit rules and citation styles (natbib vs. numeric). - Cross-Model Review: Sends the draft outline to GPT-5.5 via Codex MCP for scored feedback on logic, evidence alignment, and page feasibility. - Optional Style Reference: With --style-ref, extracts structural patterns from an exemplar paper and emits a GAP_REPORT.md flagging sections lacking evidence. - Use Case: After finishing experiments for an ICLR submission, run the skill to produce PAPER_PLAN.md with a section-by-section outline, hero figure description, and citation plan before writing any LaTeX. ## Quick Start Ask the AI to plan the paper by saying: create a paper outline for my ICLR submission using the review conclusions and experiment results in this project.
