skills

Generate Python matplotlib plotting scripts in the user's scienceplots lab style: with plt.style.context(["science", "nature"]), pparam = dict(...), raw-string LaTeX labels, and fig.savefig(..., dpi=300, bbox_inches='tight'). Support parquet, CSV, NPY/NPZ data and single-line, multi-line, scatter/errorbar, or subplot variants. Generate both the reproducible plotting script and the requested PNG/PDF/SVG figure by default. Use for publication-style plot scripts, science/nature figures, plot-from-data scaffolds, and Korean or English requests for matplotlib graph scripts.

Create rigorous teaching explanations as a single markdown document with equations, assumptions, derivations, plots, optional schematic prompts, and PDF export. Use when the user asks to explain a physics, math, ML, statistics, or CS concept; write lecture, tutorial, or seminar notes; derive results step by step; build a kind but rigorous walkthrough; or turn a paper section into classroom-ready material. Finished explanations are archived to ~/Dropbox/ConceptExplainer/<Topic>/<concept-slug>/ automatically, regardless of language.

Create or revise a markdown research or experiment report that makes the work UNDERSTANDABLE (why it matters, what the core idea is, what it achieved), backed by traceable evidence, integrated plots, optional literature support, plot manifest tracking, and version history, inside a single Harness without external Codex or Gemini calls. Use when you need to generate `report.md`, explain a result or method so a colleague can understand it, inventory or validate plots, create `plots/plot_manifest.json`, manage `report_versions.json`, connect report sections to supporting references, or turn an experiment or output directory into a reusable report workflow.

Produce a journal-club-style paper presentation (9 sections: TL;DR, Problem, Key Idea, How It Works, Key Results, Why It Matters, Strengths/Limitations/Open Questions, Discussion Questions, Takeaways) from an arXiv ID/URL, a PDF, raw text/markdown, or a local LaTeX source (.tex / project dir). Helps a reading group UNDERSTAND and DISCUSS the paper — not score or accept/reject it. Grounds every claim in the source, renders math as LaTeX, auto-matches the source language (Korean source -> Korean review). When LaTeX source is available (arXiv e-print or local .tex), embeds the paper's real figures with captions; optionally generates two friendly-whiteboard infographic figures via the bundled codex image_generation tool. Use when the user wants a journal-club review, paper walkthrough/presentation, or paper explainer, to "review this PDF/paper like a journal club", or 논문 저널클럽 리뷰/발표자료/논문 설명. For an OpenReview referee report use workshop-paper-review; for an adversarial pre-submission audit use adversarial-review.

Search the live web via z.ai web_search_prime (GLM Coding Plan MCP server). Use when the user wants current information, recent trends, news, blog posts, docs, or any web resource that academic-database search (reference-search) does not cover. Triggers on web search, 웹 검색, search the web, 최근 동향, 최신 트렌드, latest trends, find online, 인터넷에서 찾아, what's new, current state of X, search z.ai, GLM web search.

Generate a hand-drawn, whiteboard-sketch schematic that explains a concept, pipeline, architecture, algorithm, or mechanism at a glance, on a pure white background. Composes a friendly hand-lettered infographic prompt (wavy marker strokes, numbered panels flowing left to right, chunky chalk arrows, hand-written notes beside shapes) and by default renders it to a PNG via the bundled codex image_generation tool (ChatGPT OAuth); falls back to emitting the copy-paste prompt when codex is unavailable. Use when the user wants a hand-drawn / hand-written / whiteboard-style schematic, a sketch diagram that shows a schema, a friendly explainer figure, a doodle-style method overview, or 손그림/화이트보드 스타일 도식/개념도. For matplotlib data plots use scienceplot-py or xkcd-py; for a multi-style (editorial, blueprint, neon, poster) prompt-only composer use wide-slide-illustrator; for the paper-review method+results pair use journal-club-review.

Manage Vast.ai GPU cloud instances via the vastai CLI. Use this skill whenever the user mentions Vast.ai, GPU rentals, cloud GPU instances, searching for GPU offers, creating/destroying instances, vast.ai billing, or any task involving the vastai command-line tool. Also trigger when the user wants to rent GPUs, find cheap GPUs, deploy Docker containers on remote GPUs, manage remote training infrastructure, or transfer data to/from cloud GPU machines. Even if the user just says "spin up a GPU" or "find me an A100", this skill likely applies.

Fetch valid (still-open, deadline-not-passed) job postings from Academic Jobs Online (academicjobsonline.org, AJO) and the InspireHEP jobs board (inspirehep.net/jobs). Use when the user wants current academic openings: postdoc / faculty / PhD positions in physics, cosmology, HEP, ML, astrophysics or any field, filtered to postings whose application deadline has not passed. Searches both boards by default and merges the results. Manage field presets (keywords + position types + which boards), fetch and store valid postings, see what is new since last check, and inspect a posting's details. Triggers on: Academic Jobs Online, AJO, InspireHEP jobs, Inspire HEP 공고, academic job postings, postdoc openings, faculty positions, job listings, valid 공고, 학술 잡, 교수 공고, 포닥 공고, 채용 공고, 잡 마켓.

Analyze the HEP-theory postdoc rumor mill (the public Google Sheet at sites.google.com/site/postdocrumor) to see who got offers at which institutions, then pull each offer-holder's publication record from InspireHEP (and OpenAlex + Semantic Scholar for interdisciplinary people) to study what kind of profile lands where. Use when the user wants to study the HEP-theory / cosmology / hep-ph job market from the rumor mill: institute-by- institute cohort profiles (citations, papers, venues, subfield, PhD-age, named fellowships), benchmark their own InspireHEP profile against the accepted cohort, or generate a Korean job-market strategy report. Triggers on: rumor mill, postdoc rumor, HEP rumor mill, hep-th job market, theory postdoc offers, who got offers, offer holders, 루머밀, 포닥 루머, 잡마켓 분석, 오퍼 받은 사람, 합격자 실적 분석, 포닥 전략, InspireHEP 프로필 분석.

Convert Markdown to PDF using Typora's Whitey theme via pandoc + Chrome headless, replicating Typora's PDF export appearance

Read and search Proton Mail through the local Proton Bridge (IMAP over STARTTLS on 127.0.0.1). Use when the user mentions Proton Mail, proton-bridge, asks to check/list/search their Proton inbox, read a specific Proton email, or list unread mail. Read-only: it does not send, delete, or move mail. Credentials live at `~/.proton-imap`.

Generate a Python matplotlib plot script following the user's mandatory xkcd style — `with plt.xkcd():` context, `pparam = dict(...)` axis-config, raw-string LaTeX labels, wide canvas (`figsize=(10, 6)`), `dpi=300` savefig. Supports parquet / CSV / NumPy (`.npy`, `.npz`) and four variants: single line, multi-line + legend, scatter / errorbar, multi-panel subplots. Writes the `.py` only — does NOT execute it; the user runs it (typically `uv run`). Use when asked to write an xkcd-style / hand-drawn / sketch-style matplotlib script, plot data in xkcd style from parquet / CSV / `.npy`, scaffold an xkcd line / scatter / errorbar / subplot script, or match the lab's xkcd template. Triggers on: xkcd plot, xkcd matplotlib, xkcd style, hand-drawn / sketch plot, plt.xkcd, comic-style plot; xkcd 플롯, xkcd 그래프, 손그림 그래프, 스케치 스타일 플롯, xkcd 스크립트, xkcd 코드.

Maintain a research log that acts as an advisor, not just a diary. Registers projects; records decisions with generalizable lesson extraction; saves session state; recalls cross-project lessons and prior solutions when stuck; checks proposed approaches against promoted anti-pattern rules; runs weekly review with lesson promotion and dashboard refresh. Use when the user asks to initialize a research-log project, record an experiment or decision, save session state, recall past lessons or how a blocker was solved before, check a planned approach against past mistakes, or review all projects.

Back up untracked research report directories (outputs/, results/, report(s)/) from local projects into the locally synced Dropbox folder, organized by category/project. Discovers candidate directories that git does not track, keeps a registry under ~/.config/research-backup/, and mirrors additively with rsync (deletions are never propagated). Use when the user asks to back up research outputs or reports to Dropbox, register new report directories for backup, check what would change before backing up, or says backup outputs, research backup, outputs 백업, 결과 백업, 리포트 백업, 연구 백업, Dropbox 백업 모음. For single-file upload/download/share through the Dropbox API use the dropbox skill instead.

Search and curate academic references for a topic, claim, section, baseline, method, evaluation need, or report section using domain-aware routing across InspireHEP, OpenAlex, and Semantic Scholar, with markdown-ready summaries and optional save paths. Use when the user asks for papers, references, citations, related work, supporting literature, survey papers, background sources, baseline citations, evidence to support a report or research claim, or wants section-by-section citation help while drafting `report.md`.

Build and manage a local MkDocs Material "research portal" over a folder of Typora/markdown research notes (e.g. ~/notes or a Dropbox-synced folder), keeping the original .md files untouched. Use this whenever the user wants to collect, browse, organize, or curate a folder of dated markdown notes into a searchable navigable site; wants a lighter local alternative to Obsidian / Notion / Outline / Logseq for their own notes; needs perfect LaTeX/math rendering in a notes site; wants notes grouped by project in a sidebar; or wants to tag/rename notes by project without breaking image links. Trigger even if they don't say "MkDocs" or "portal" but describe wanting to "정리해서 모아 보기", make a personal wiki, or render math from their note folder.

Produce OpenReview-ready peer reviews for ICML/NeurIPS/ICLR workshop papers (4-8 page submissions). Covers PDF intake, evidence-grounded review drafting, anti-anchoring scoring calibration, parallel fact-check verification against the source PDF, AI-writing-pattern removal, and bilingual (Korean draft → English submission) workflow. Produces a tight ~400-word review with Title, Rating, Confidence, Summary, Strengths, Weaknesses, and Questions for Authors — without an Overall section. Use when the user mentions reviewing a workshop paper, drafting an OpenReview submission, preparing a referee report for ICML/NeurIPS/ICLR workshops, fact-checking a review draft, or translating a Korean review draft into a publication-ready English submission.

Generate BibTeX entries for academic references from arXiv IDs, DOIs, titles, URLs, batch lists, or reference-search JSON. Route HEP papers to InspireHEP, non-HEP papers to Google Scholar when available, and CrossRef DOI BibTeX as a fallback; preserve source-native keys and print or append to a .bib file. Use when the user asks for bibtex, citation entries, bibliography filling, arXiv/DOI/title-to-BibTeX, InspireHEP BibTeX, publisher BibTeX, or converting discovered references into a .bib file.

Bilingual section-by-section workflow for Overleaf physics-paper drafts: Korean draft, user iteration, Opus-direct English LaTeX translation, out-of-tree build, and sync hygiene. Use when starting, iterating, translating, or finalizing an Overleaf paper section; merging Korean drafts into main.tex; verifying citations; enforcing no em/en dashes, no unsupported novelty claims, and no forward references in background sections; or building JCAP/PRD-style sections with figures and BibTeX.

Compose image-generation prompts for wide 18:9 infographic slides that explain methodology, data pipelines, architectures, or research ideas. Support Friendly Whiteboard, Editorial Magazine, Engineering Blueprint, Swiss Minimalist, Dark Tech/Neon, and Scientific Poster styles, with English-only on-image text and style-specific hard negatives. Use when the user asks for slide illustrations, pipeline diagrams, paper hero figures, journal-style methodology figures, keynote backdrops, or prompts for ChatGPT Image, gpt-image, DALL-E, Sora, or Midjourney.

Bidirectional real-time sync between an Overleaf project and a local directory using the `overleap` Node.js CLI. Use when the user mentions Overleaf, overleap, "sync my Overleaf project", "edit Overleaf locally", "pull / push Overleaf", "compile Overleaf to PDF from CLI", or asks Claude to edit `.tex` files that should reflect on Overleaf in real time. Also use when the user wants to set up or refresh their Overleaf session cookie for the CLI. Authentication uses an `OVERLEAF_COOKIE` value, stored in a per-project `.env` file or in the environment.

Manage calendars, events, tasks, and tags through the Morgen API (docs.morgen.so). Use when the user mentions Morgen, asks to list/create/update/delete calendar events or tasks across Google Calendar, Microsoft 365, iCloud, Fastmail, or CalDAV accounts connected through Morgen, or wants to query connected integrations and tags. Credentials live at `~/.config/morgen-skill/credentials.json`.

Run an adversarial peer-review swarm on a paper draft, report, or manuscript by spawning parallel persona subagents (hostile theorist, experimentalist, statistician, journal editor, citation auditor, figure critic) and synthesizing their critiques into a ranked fix list with suggested defense experiments. Use when the user asks for a paper review, referee simulation, desk-reject check, pre-submission audit, citation audit, figure audit, novelty attack, or wants feedback on a draft report before submission or internal circulation.

Classify uncommitted changes into commit-ready, failure-archive, and ambiguous buckets, then produce clean grouped commits with no co-author attribution. Use when the user asks to commit, tidy up `git status`, prepare a PR, archive failed experiments to a `failure/` folder, separate research artifacts from code before a commit, or clean up a noisy working tree at the end of a session.

Upload files to Dropbox, download from Dropbox, and create or reuse shared links. Use when the user mentions Dropbox, asks to upload/download/share a file via Dropbox, or wants a shareable link for a file already in Dropbox. Use `~/.config/dropbox-skill/credentials.json` as the only supported credentials path.

Generate publication-quality academic diagrams and statistical plots from text descriptions using the paperbanana CLI. Supports methodology diagrams, statistical plots, diagram evaluation, and iterative refinement with user feedback. Use when the user asks to: create a diagram, generate a figure, make a plot from data, evaluate diagram quality, refine/improve a generated diagram, or anything related to academic illustration generation. Triggers on: "diagram", "figure", "plot", "paperbanana", "academic illustration", "methodology diagram", "evaluate diagram", "refine diagram", "improve figure".