What problem does it solve? Academic writing with AI often produces text disconnected from real evidence: citations that were never read, claims unsupported by full texts, and drafts with no verifiable quality gates. This Skill connects topic definition, literature search, PDF download, Zotero organization, evidence-grounded writing, figure generation, citation audit, and conservative polishing into one traceable workflow where every claim maps to graded evidence. ## Core Features & Use Cases - Eight-step workflow with direct entry: Enter at any step (topic, outline, search plan, search & scoring, PDF download, Zotero alignment, writing & audit, polishing) based on materials you already have, without rerunning prior steps. - Evidence grading and quality gates: Distinguishes metadata-only, abstract-only, and full-text evidence; weak evidence cannot support strong claims, and completion gates block premature "done" declarations. - Literature retrieval and download routing: Multi-source search (OpenAlex, Crossref, Semantic Scholar, PubMed, arXiv, CNKI, Wanfang) with scoring tiers, plus PDF download routing via OA resolvers, Sci-Hub, and CDP browser automation with login checkpoints. - Writing, figures, and audit: Step 7 generates evidence-mapped drafts, native paper flowcharts/architecture diagrams as semantic SVG with black-and-white publication mode, scientific figure reproduction, and per-claim citation auditing. - Use Case: A graduate student with a Zotero library and partial draft enters directly at Step 7, builds an evidence map and argument plan, writes chapters with audited citations, generates publication-ready black-and-white diagrams, then runs Step 8 conservative polishing with an AI-trace diagnostic report. ## Quick Start Use more-paper-workflow and go directly to Step 7 with my Zotero library and draft, building the evidence map and argument plan before writing the target section.
