VELA

Use when the user asks to reduce AI-like academic prose, lower AIGC-like surface signals, humanize, deslop, remove translationese, or make Chinese or English manuscript text sound more authorial while preserving facts, citations, data, results, and uncertainty.

Use when the user asks to review, critique, summarize, or assess one academic paper, preprint, or research article, especially for referee reports, methodology checks, contribution assessment, top-journal internal review, or social-science-style evidence review.

Use when the user needs top-journal-quality social-science statistical figures and tables: regression tables, descriptive tables, coefficient plots, event-study plots, predicted probabilities, marginal effects, robustness figures, text-analysis summary charts, network metrics charts, publication-ready captions, or figure/table consistency review. Not for conceptual mechanism diagrams or image2-first scientific illustrations; use research-figure-studio for those.

Use when the user asks to draft, rewrite, polish, translate, or final-check academic manuscript text, including title, abstract, introduction, literature review, methods, results, discussion, conclusion, Chinese/English polishing, or journal-style manuscript revision. This skill distinguishes Chinese vs English journal writing and social-science vs technical manuscript conventions before editing.

Use when the user is doing research, literature review, computational social science, writing, or platform-case work and wants Codex to choose the right skills, MCP profile, plugins, and compute target automatically while explaining the route first.

Use when the user intends to create, redesign, polish, audit, or submit social-science research figures, paper figures, scientific illustrations, mechanism diagrams, conceptual frameworks, workflow diagrams, research-design figures, result figures, multi-panel figures, or image2-assisted academic visuals. Automatically use this skill for requests to draw figures for papers, reports, manuscripts, top-journal review, or research presentations.

Use when the user needs to answer reviewer comments, build a response matrix, classify revision requests, audit a response letter, verify a revised manuscript against review comments, or prepare a revise-and-resubmit package.

Use when a social-science paper project needs a submission-ready package, especially with manuscript, figures, tables, citations, reviewer comments, data/code notes, reproducibility materials, and final freeze checks.

Use when the user asks to clean, organize, audit, hand off, archive, or close a project folder, or before finishing substantial file-generating work, to classify temporary files, dead files, duplicate drafts, stale outputs, and misplaced artifacts while preserving source, evidence, data, logs, and user-owned materials.

Use when the task is to obtain and verify full-text PDFs for a reference list or literature corpus: OA lookup, Google Scholar right-side PDF/all-versions clicking, CNKI authorized download, user-provided PDFs, acquired/missing lists, and Zotero attachment planning. Not for final writing citation capture.

Priority evidence-checking coordinator for structured reading and literature verification. Use first when the user says 开始结构性阅读、结构化阅读、结构性读文献、筛选参考文献、候选文献表、文献核验、证据核验、引文核验、引用核验、引文证据、引用证据、证据句、支撑判断 or 证据包, or asks to verify whether cited papers support manuscript claims, mark evidence sentences in PDFs, build a structured reading table, validate acquired full texts, build an about-50-paper candidate pool after a research framework or empirical/experimental results, write an evidence-based literature review, or produce auditable citation-evidence packages for social-science manuscripts. Coordinates reference-fulltext-acquisition, citation-verifier, systematic-literature-review, writing-reference-capture, pdf, and Zotero rather than replacing them.

Use when tasks involve reading, creating, or reviewing PDF files where rendering and layout matter; prefer visual checks by rendering pages (Poppler) and use Python tools such as `reportlab`, `pdfplumber`, and `pypdf` for generation and extraction.

Use when the user needs quantitative social-science analysis, empirical modeling, causal inference, regression, panel models, DID, IV, RDD, PSM, SCM, DML, robustness checks, empirical tables, figure-ready model outputs, or replication-package preflight.

Use when the user wants to create, redesign, polish, audit, or export research presentations, PPT, PPTX, slides, deck, PowerPoint, Google Slides, defense slides, seminar talks, project reports, paper-to-slides, visual briefings, or web-based horizontal presentation pages. For HTML web decks, delegate visual execution to the installed original guizang-ppt-skill.

Use when a research task has entered project-type multi-agent planning and Codex must turn the selected route into a concrete clarification card, agent set, target-specific review mapping, dispatch artifact, and canonical project skeleton without breaking the V6.4 contract.

Use when Codex needs to inspect Douyin, Bilibili, WeChat public-article pages, boards, favorites, collections, posts, videos, or similar platform artifacts, and must extract only directly visible evidence from a logged-in browser or a stronger installed backend.

Use when a manuscript, review response, report, or export is being delivered and only the papers actually cited or used in that writing need citation-evidence verification, Zotero entry, and Obsidian sync. DOI must be verified when present in the evidence; user-provided sources, complete paper PDFs, and public academic index records can waive the requirement that every formal reference must have a DOI. Not for downloading PDFs or building a literature corpus.

Use this when checking, weakening, splitting, or linking VELA claims against named evidence records.

Use this before sharing VELA deliverables to check unsupported claims, private material, and validation gaps.

Use this when promoting VELA materials into evidence records with source, access, verification, and rights notes.

Use this when turning a user request into a bounded VELA Codex handoff packet.

Use this when adding collected source material to a VELA project. Keep material separate from verified evidence.