Plugin Claude Code
METASCI — a personal Claude Code skill collection spanning academic research (knowledge/logic/vocabulary extraction, writing, rewriting, multi-reviewer proofreading, agentic research) and dialogue personas (Virgil, Beatrice, Socrates, Picasso, Feynman digest). Supports AI collaboration based on Gemini CLI.
Instructions file
Claude Code instructions for kkhgeo/METASCI, covering metasci — claude code project instructions, project overview, skill architecture, per-runtime builds and installing on a machine.
Skill Claude CodeCodex
Extract core knowledge from paper PDFs and save as structured markdown. Classify into 5 epistemological categories (Theory, Empirical, Methodology, Context, Critique). Supports parallel processing via Subagent. Automatically invoked on requests like "논문 분석", "지식 추출", "Knowledge 저장".
Skill Claude CodeCodex
Systematically extract structure (sections/paragraphs), argument logic (inter-paragraph and intra-paragraph), and sentence frames (rhetorical templates) from academic paper PDFs. Saves results as structured markdown. Triggers: "논리 분석", "구조 분석", "문장 형식 추출", "논문 구조 파악", "logic analysis", "structure analysis", "sentence…
Skill Claude CodeCodex
Build a reusable STYLE CORPUS from reference papers — one folder per paper holding normalized text, per-section splits, an exhaustive sentence-frame catalog, a measured style vocabulary, a lean Style Card, and a machine-readable manifest. Runs three lenses over shared, pre-normalized text — A (architecture + sentence…
Skill Claude CodeCodex
Exhaustively extract ALL content words from academic paper PDFs, organized by section and part of speech. Technical/domain terms are separately identified and classified. Triggers: "단어 추출", "어휘 추출", "용어 추출", "품사별 추출", "vocabulary extraction", "term extraction".
Skill Claude CodeCodex
Use when an academic draft in English — a paragraph, section, or report chapter, including a bilingual working document whose manuscript prose is English — must be revised to match reference papers whose style has already been extracted into an extraction-style corpus. Triggers: "문체 교정", "스타일 맞춰줘", "톤 맞춰줘", "스타일 적용"…
Skill Claude CodeCodex
Plan, write, finalize, and review research presentation structure and slide content. Use when Codex needs to turn papers, reports, existing decks, notes, or dense source material into an approved slide content specification with deck promise, audience context, narrative spine, slide topics, final titles, lead…
Skill Claude CodeCodex
Explore, design, implement, and visually review Korean research slide decks in the Onto house style. Use when Codex needs early design direction options, Onto design tokens, Paperlogy typography, HTML slide implementation, cover/section/overview/method layouts, PDF export, visual redesign, screenshot verification, or…
Skill Claude CodeCodex
Install the author-dialogue communication rules (AI-tell removal, adapted from pstack's unslop) into a manuscript workspace as one standalone AGENTcommunication.md, so CLAUDE.md and AGENTS.md stay short and the rules survive a workspace regeneration. Use when the user says "대화 규율 추가해줘", "대화 규칙 파일 만들어줘", "unslop 규칙…
Skill Claude CodeCodex
Install workspace-level figure rules for publication figures that go from Python (matplotlib) to Illustrator with minimal post-editing: drops FIGURES.md (drafting and audit conventions), figurespec.yaml (every number in pt/mm, Nature defaults), and figspec.py (library that applies the spec, prints a parameter sheet…
Skill Claude CodeCodex
Initialize an academic writing/manuscript project folder by generating CLAUDE.md and AGENTS.md instruction files that configure any AI agent (Codex CLI, Codex, Cursor, etc.) as a scholarly manuscript reviewer rather than a coding agent. Use this whenever the user starts a new writing or manuscript project, asks to…
Skill Claude CodeCodex
Meta-workflow skill for designing, generating, and evaluating Introduction sections of environmental science papers. Builds a structural Blueprint from Results/Discussion or research topic, writes publication-ready Introductions using Knowledge files (literature), and evaluates quality using the CARS model. Activate…
Skill Claude CodeCodex
Post-hoc quantitative correction of a publication figure drawn under FIGURES.md / figurespec.yaml / figspec.py (installed by init-figure-rules). Renders, audits the exported PDF (canvas mm, every font size and stroke width in pt, text kept as editable text not outlines, font family, panel groups), detects overlapping…
Skill Claude CodeCodex
Check whether distinctive English academic phrases and collocations are attested in published literature, with traceable examples and verified alternatives. Use for "용례 확인", "이 표현 실제 논문에 쓰였어?", "표현 검증", "phrase check", or "is this used in papers". Do not use for generic proofreading, argument or style review…
Agent
A search-and-merge agent that independently looks up 14 expressions and returns only the combined structured result.
Skill Claude CodeCodex
A Korean-language proofreading process for academic and policy writing that uses several review levels. It makes definitive decisions about sentence and notation rules, uses panel voting for argument structure, and leaves style choices to the author.
Agent
Verify the load-bearing facts of a reviewed paragraph.
Agent
A decision-making agent for the first and second levels of a review process. It compares the original with the changed version, checks the first level, and makes a majority decision at the second level.
Agent
A review agent used only for the second and third levels of a review process. The first level is handled directly by the coordinator using fixed rules.
Skill Claude CodeCodex
Panel proofreading of an existing draft — several reviewers with different knowledge allocations review the same text in parallel, a judge scores their competing rewrites, and the original can win. Covers a full draft, one section, or one paragraph down to the sentence. Invoke only when the user names the panel: "메타…
Agent
Build the Knowledge Bank by discovering, indexing, parsing, and distributing knowledge files to reviewers. Dual mode: local file scanner + web search supplement.
Skill Claude CodeCodex
Review and improve user's draft (paragraph/section) against multiple reference paper extraction files (logic + vocab). Per-paper Reviewer subagents evaluate the draft using four academic writing principles (with extraction files as evidence), then Orchestrator synthesizes findings and produces a holistic rewrite.…
Skill Claude CodeCodex
A writing-editing skill that detects and revises patterns commonly associated with AI-written academic prose. It checks vocabulary, sentence structure, flow, formatting, and changes made during revision.