Communitygithub.com

MisterBrookT/paper-skills

Journal-ready AI skills for every research discipline.

paper-skills とは?

paper-skills is a Claude Code agent skill that journal-ready AI skills for every research discipline.

対応~Claude Code~Codex CLI~Cursor
npx skills add MisterBrookT/paper-skills

お気に入りのAIに質問する

このエージェントスキルを事前に読み込んだ状態で新しいチャットを開きます。

ドキュメント

Plot Skill

plot is the original flagship skill in Paper Skills. It focuses on paper-ready scientific figures across disciplines, not generic chart styling.

Core Scope

Use this skill when the user needs:

  • a discipline-specific research figure;
  • an editable scientific schematic or diagram;
  • a figure generated from domain materials such as netlists, paper text, pathway notes, system descriptions, time-series events, or rough sketches;
  • an expert-review rubric for what makes a figure acceptable in a field;
  • a benchmark case for domain-aware figure generation.

Non-Goals

  • Do not claim to replace professional EDA, CAD, statistical, or simulation tools.
  • Do not optimize only for visual prettiness.
  • Do not turn field-specific figures into generic PowerPoint-style diagrams.
  • Do not invent scientific content not present in the user's materials.

Workflow

  1. Identify domain and figure family.
  2. Collect source materials and constraints.
  3. Read the matching domain note under domains/ when available.
  4. Convert inputs into an auditable drawing spec.
  5. Produce a baseline figure in an editable format when possible.
  6. Ask for or infer domain-specific failure criteria.
  7. Distill failures into rubric items and skill rules.
  8. Re-run generation with the updated rules.
  9. Save benchmark artifacts when the task is part of evaluation.

Domain Contract

Before generation, write a short contract:

  • domain;
  • figure family;
  • source inputs;
  • target output format;
  • must-preserve entities and relationships;
  • hard-fail criteria;
  • stylistic conventions;
  • evaluation plan.

Output Preference

Prefer editable outputs over flat images:

  1. .drawio, SVG, Mermaid, TikZ, matplotlib source, or structured JSON;
  2. PDF/PNG export for manuscript insertion;
  3. flat raster only when no editable route is available.

First Domain: Circuit Figures

Initial seed domains live under:

  • domains/integrated-circuits/
  • domains/computer-science/

Integrated-circuit seed family:

transistor-level analog schematic illustration

Goal: turn existing circuit materials into paper-ready schematic illustrations. Inputs may include SPICE/netlist, rough EDA screenshots, paper text, hand sketches, component tables, and visual references.

The target is not circuit design automation. The target is scientific figure automation for readable, editable, publication-ready circuit illustrations.

Second Domain: Computer Science Figures

Computer-science seed families:

  • system diagrams;
  • model architecture diagrams;
  • pipeline and dataflow figures;
  • evaluation setup figures;
  • benchmark result figures when ordinary charts are not enough.

The target is not generic slide diagrams. The target is paper-ready CS figures with clear modules, data flow, experimental structure, and reproducible editing handles.

関連スキル