Haojae/scipilot-figure-skill

Última actividad de origen registrada: 15 de junio de 2026 a las 03:39

Catálogo de SkillsMP actualizado: 2 de noviembre de 2026 a las 18:02

skills recopiladas: 1

Estrellas en GitHub: 2757
Forks en GitHub: 100

Instalar con un asistente de IA

Copia este prompt en el asistente de IA que estés usando.

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/Haojae/scipilot-figure-skill.

README del repositorio

Archivo de origen del README: README.md

Leer el README original

Mostrar README completo

# scipilot-figure-skill > SciPilot Skills family. Scientific data **visualization advisor** — thinks first, plots second. > SciPilot Skills 家族成员 — 科研数据**可视化顾问**，先思考后绘制。 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE) [![Python: 3.9+](https://img.shields.io/badge/Python-3.9%2B-3776AB.svg)](#dependencies--依赖) [![Status: v2.1.0](https://img.shields.io/badge/Status-v2.1.0-success.svg)](#) [![Advisor Mode](https://img.shields.io/badge/Mode-Advisor%2BPlotter-c41e3a.svg)](#为什么这不只是个画图工具) [![Stack](https://img.shields.io/badge/Stack-matplotlib%20%7C%20seaborn%20%7C%20plotly-orange.svg)](#) [![Claude Code Skill](https://img.shields.io/badge/Claude%20Code-Skill-orange.svg)](https://claude.com/claude-code) A [Claude Code](https://claude.com/claude-code) / [Codex](https://github.com/openai/codex) / Cursor Skill that does **two things in order**: (1) profiles your data and recommends the right chart for the argument you want to make, (2) renders it at publication grade for Nature / Science / IEEE / Elsevier / PNAS / Chinese journals. Built on **matplotlib + seaborn + SciencePlots** (static) and **plotly** (interactive), with **CJK font auto-configuration** so Chinese text never renders as boxes. > [中文文档](#中文文档) | [English](#english) --- ## 中文文档 ### 概览科研工作者最大的画图痛点往往不是"不会用 matplotlib"，而是"手上一堆数据，不知道该用什么图把结论讲清楚"。`scipilot-figure-skill` 是 SciPilot Skills 家族的第二个成员，专做这件事—— **它的首要能力是【思考与判断】，其次才是【绘制】。** ### 为什么这不只是个画图工具 ``` 普通画图工具 scipilot-figure-skill ───────────── ───────────────────── 你说 "画柱状图" → 先 EDA：每列类型、样本量、分布、异常值、相关性直接 plt.bar() 再问："你想论证什么？组间差异？关系？趋势？" 按数据特征 + 论证目标查决策框架推荐图型 n=5 想画均值柱时主动拦截 → 改用 stripplot 维度 > 12 时建议拆图，不硬塞最后才进入绘制 → 期刊规范 → 自检 → 导出 ``` ### 核心工作流（8 步） ``` 0. 理解任务 ── 这张图要论证什么？数据在哪？ ↓ 1. 剖析数据 ── profile_data.py：列类型/样本量/分布/异常/相关 ↓ 2. 选图 ── chart_selection.md：按数据形态+论证目标决策 ↓ (主动拦截 → viz_pitfalls.md) 3. 查规范 ── journal_specs.md：栏宽/字号/DPI/字体 ↓ 4. 配环境 ── setup_style.py：期刊预设+CJK字体 ↓ 5. 绘制 ── plot_recipes.md：9 类图配方 ↓ 6. 自检闭环 ── 程序自检 visual_qa + AI 读图 visual_review.md（缺字/裁切/重叠/对齐→回改） ↓ 7. 导出 ── export_figure.py：多格式+按最终尺寸+灰度预览 ``` ### v2.1 新增：出图后的视觉自检闭环普通画图工具画完就结束——没人回看成图，于是中文方框、文字被裁、图例压数据、子图编号乱放，全留到投稿才暴露。v2.1 补上**出图后的闭环**： ``` 绘制 → 渲染 PNG → 程序自检(缺字/裁切/重叠) + AI 读图(遮盖/子图对齐/灰度) ↓ 发现问题回到对应步骤改图 → 重渲 → 再读，直到通过 ``` - **程序抓确定性问题**：`visual_qa.audit_layout()` 同时盯 warning + logging 两条通道拦截缺字乱码，并查文字越界裁切、刻度标签重叠。 - **AI 读图抓感知性问题**：渲成 PNG 后用多模态读图能力核对图例是否压数据、子图 a/b/c 是否对齐、配色灰度可分——这些程序查不出。 - **子图标签一行对齐**：`layout_tools.add_panel_labels(fig)` 用统一 figure 坐标让 a/b/c/d 横竖成线，不再手摆易错位。 - **字体兜底**：`setup_style` 全模式默认修负号方框 + 自动配 CJK 字体，从源头少出乱码。详见 [`references/visual_review.md`](references/visual_review.md)。 ### 五条硬性原则 1. **按最终尺寸出图不二次缩放** — `figsize=(3.5, 2.625)` 直接定 Nature 单栏 2. **矢量优先** — 数据图走 PDF / SVG / EPS；照片才用 TIFF/PNG；**绝不 JPEG** 3. **配色对色盲友好** — 默认 Okabe-Ito，加冗余编码，出图前查灰度 4. **字号在最终尺寸下可读** — 7-9 pt，最小 6 pt 5. **误差必有交代** — 图注必须写清 SD/SEM/95%CI + n + 检验方法 ### 主动拦截的画图错误（精选）完整 15 条在 [`references/viz_pitfalls.md`](references/viz_pitfalls.md)。 - **P1**：n<10/组的均值柱掩盖分布 → 改箱线 + stripplot - **P2**：双 Y 轴的"相关"是作图者捏造的 → 拆子图或标准化 - **P3**：饼图与 3D 图 → 横向柱状 - **P4**：Y 轴不当截断 → 从 0 起或断裂明示 - **P6**：x 是分类用折线连均值 → 散点 / 柱状 - **P14**：rainbow / jet 色图 → viridis / RdBu_r - **P12**：一图多论点 → 拆图，一图一结论 - **P16**：中文/负号变方框 → setup_style 配 CJK + 关 unicode_minus，导出前 visual_qa 拦截 - **P18**：子图 a/b/c 乱放 → add_panel_labels 统一对齐 ### 安装 ``` 请帮我安装这个 Skill：https://github.com/Haojae/scipilot-figure-skill.git ``` 或手动： ```bash git clone https://github.com/Haojae/scipilot-figure-skill.git \ ~/.claude/skills/scipilot-figure-skill pip install -r ~/.claude/skills/scipilot-figure-skill/requirements.txt ``` `SciencePlots` / `pypdf` / `kaleido` 是可选增强，缺失时不影响运行。 ### 中文支持 `setup_style(lang='zh')` 按优先级查找： ``` Noto Sans CJK SC > Source Han Sans SC > SimHei > Microsoft YaHei ``` 找不到任何 CJK 字体时抛出清晰的安装提示。中文期刊"宋体 + Times New Roman 数字"混排：传 `serif_for_zh=True`。 ### 使用示例（4 个，**至少 2 个体现"思考"**） **例 1：用户只丢一个 CSV** ``` 我有这份 results.csv，帮我画成论文图。 ``` → Skill 不会直接画。会先： 1. 问"你这份数据主要想说服读者相信什么？" 2. 跑 `profile_data.py` 输出剖析报告 3. 给出推荐图型 + 理由 + 备选 4. 用户确认后才进入绘制流程 **例 2：用户指定的图型不合适** ``` 帮我画 3 组各 5 个样本的均值柱状图。 ``` → Skill 主动拦截： > n=5 太小，均值柱掩盖分布——审稿人很可能要求"show individual points"。 > 我建议改成 **箱线 + stripplot 叠加每个点**：5 个点直接可见。要按原方案画，还是改？ **例 3：多面板组合** ``` 给我画 Figure 1：4 panel，分别是 PCA、loss、混淆矩阵、生存曲线。 Nature 双栏。 ``` → 各 panel 字号配色统一，子图标签 a/b/c/d 加粗左上，导出 PDF + 灰度预览。 **例 4：带显著性标注** ``` 3 组数据画箱线 + 显著性。 ``` → 先确认 n、检验方法、是否多重校正；图注必须写清统计内容。 ### 命令行直接调脚本 ```bash # 数据剖析（思考的起点） python scripts/profile_data.py results.csv --group group --group condition # 列可用 CJK 字体 python scripts/setup_style.py --list-fonts # 演示导出 python scripts/export_figure.py demo --out ./test_demo # 合规自检 python scripts/check_figure.py figs/*.pdf --min-dpi 300 --strict ``` ### SciPilot Skills 家族 | Skill | 状态 | 功能 | |---|---|---| | scipilot-cite-skill | [v1.0.0](https://github.com/Haojae/scipilot-cite-skill) | 文献检索与引用插入 | | **scipilot-figure-skill** | **v2.1.0 (本仓库)** | **可视化顾问 + 绘制 + 视觉自检闭环** | | scipilot-writing-skill | [v1.0.0](https://github.com/Haojae/scipilot-writing-skill) | 学术写作与润色 | | scipilot-review-skill | 规划中 | AI 模拟审稿 | | scipilot-submit-skill | 规划中 | 投稿格式适配 | | scipilot-read-skill | 规划中 | 论文阅读与翻译 | ### 许可证 [MIT](LICENSE) © 2026 Haojae --- ## English ### Overview The hardest part of scientific plotting is rarely "I don't know matplotlib" — it's "I have data, I don't know which chart conveys my conclusion." `scipilot-figure-skill` is the second member of the SciPilot Skills family, built around this insight: **think first, plot second**. ### Why this isn't just a plotting tool ``` Generic plotter scipilot-figure-skill ────────────── ────────────────────── "plot a bar chart" → Profiles data first: types, n, distribution, outliers plt.bar() Asks: "What argument do you want this figure to make?" Decides chart type from data shape + intent Refuses bad choices (n=5 mean bar → stripplot instead) Suggests splitting figures when dimensions exceed 12 Only then renders → spec → audit → export ``` ### 8-step core workflow ``` 0. Understand — what does this figure argue? where is the data? ↓ 1. Profile — profile_data.py: types / n / distribution / outliers / corr ↓ 2. Select — chart_selection.md: decision framework by shape + intent ↓ (active interception → viz_pitfalls.md) 3. Spec — journal_specs.md: column width / font / DPI ↓ 4. Style — setup_style.py: journal preset + CJK font config ↓ 5. Plot — plot_recipes.md: 9 recipe families ↓ 6. Self-check — visual_qa (program) + AI reads the PNG (visual_review.md): glyphs/clipping/overlap/alignment ↓ 7. Export — export_figure.py: multi-format + final size + grayscale ``` ### New in v2.1: a post-render visual self-check loop Generic plotters stop at "saved" — nobody looks at the result, so CJK tofu boxes, clipped labels, legends covering data, and misaligned panel letters all survive to submission. v2.1 closes the loop **after rendering**: ``` plot → render PNG → program audit (glyphs/clipping/overlap) + AI reads image (occlusion/panel alignment/grayscale) ↓ issues found go back, fix, re-render, re-read — until clean ``` - **Program catches deterministic issues**: `visual_qa.audit_layout()` traps missing-glyph warnings (both warnings and logging channels), text clipped off-canvas, and overlapping tick labels. - **AI catches perceptual issues**: after rasterizing to PNG, the skill reads the image to check whether the legend covers data, whether panel letters a/b/c line up, and whether colors stay distinct in grayscale. - **One-call panel alignment**: `layout_tools.add_panel_labels(fig)` places a/b/c/d in unified figure coordinates so they align both ways. - **Font safety net**: `setup_style` fixes the minus-sign box in all modes and auto-configures CJK fonts. See [`references/visual_review.md`](references/visual_review.md). ### Five hard rules 1. **Render at final size, never rescale** — set `figsize=(3.5, 2.625)` directly 2. **Vectors first** — PDF / SVG / EPS for data figures; **never JPEG** 3. **Colorblind-safe palette** — Okabe-Ito default + redundant encoding 4. **Readable type at final size** — 7-9 pt body, 6 pt minimum 5. **Errors must be explained** — captions state SD/SEM/CI + n + test type ### Actively intercepted mistakes Full 15 in [`references/viz_pitfalls.md`](references/viz_pitfalls.md). - **P1**: mean-only bar charts with n<10 hide distributions → use box + stripplot - **P2**: dual-Y axis fabricates correlations → split panels or standardize - **P3**: pie charts and 3D bars → horizontal bar - **P14**: rainbow / jet colormaps → viridis / RdBu_r - **P12**: one figure with five points → split, one figure per claim - **P16**: CJK / minus-sign tofu boxes → setup_style configures CJK + disables unicode_minus; visual_qa blocks leftover missing glyphs - **P18**: scattered panel letters → add_panel_labels aligns a/b/c in one call ### Installation ``` Please install this Skill for me: https://github.com/Haojae/scipilot-figure-skill.git ``` Manual: ```bash git clone https://github.com/Haojae/scipilot-figure-skill.git \ ~/.claude/skills/scipilot-figure-skill pip install -r ~/.claude/skills/scipilot-figure-skill/requirements.txt ``` ### Usage examples **Example 1: bare CSV** ``` I have results.csv. Plot it for my paper. ``` The skill will: profile the data, ask what argument the figure should make, recommend a chart with rationale + alternatives, and only then plot. **Example 2: bad request, gentle refusal** ``` Plot a bar chart of group means: 3 groups, 5 samples each. ``` The skill will reply: > With n=5 per group a mean bar chart hides the distribution and reviewers > typically ask for individual data points. I recommend **box plot + > stripplot overlay** — all five points are visible and the distribution > is clear. Stick with bars, or switch? **Example 3: multi-panel composite** ``` Figure 1, 4 panels: PCA, loss curves, confusion matrix, survival. Nature double column. ``` **Example 4: statistical comparison** ``` 3 groups, box plot with significance. ``` ### SciPilot Skills family | Skill | Status | Purpose | |---|---|---| | scipilot-cite-skill | [v1.0.0](https://github.com/Haojae/scipilot-cite-skill) | Reference discovery & insertion | | **scipilot-figure-skill** | **v2.1.0 (this repo)** | **Advisor + renderer + visual self-check loop** | | scipilot-writing-skill | [v1.0.0](https://github.com/Haojae/scipilot-writing-skill) | Writing & polishing | | scipilot-review-skill | Planned | AI peer-review simulation | | scipilot-submit-skill | Planned | Submission formatting | | scipilot-read-skill | Planned | Paper reading & translation | ### License [MIT](LICENSE) © 2026 Haojae ### Dependencies / 依赖 ``` matplotlib>=3.7 seaborn>=0.13 plotly>=5.18 pandas>=2.0 numpy>=1.24 scipy>=1.10 Pillow>=10.0 SciencePlots>=2.1 # optional pypdf>=4.0 # optional kaleido>=0.2.1 # optional PyMuPDF>=1.23 # optional; only for visual_qa.render_preview() of saved PDFs ``` Python 3.9+ recommended.

Cargando el README del repositorio…

Skills en este repositorio

Científicos biológicos, todos los demás

1 categorías ocupacionales · 100% clasificado

skill

ocupación

descripción

actualizado

scipilot-figure-skill

Científicos biológicos, todos los demás

SciPilot Skills 家族成员，负责科研数据可视化——但定位不是"画图工具"，而是"可视化顾问"。先做数据剖析（列类型/样本量/分布/异常值/分组结构/相关性），再结合用户的论证目标推荐图型，主动拦截科研画图的经典错误（小样本画均值柱掩盖分布、双 Y 轴、饼图、Y 轴不当截断、rainbow 色图、把分类点连成折线等），最后产出 Nature / Science / IEEE / Elsevier / PNAS / 中文核心期刊级别的成图。覆盖纯数据图：折线、柱状、散点、箱线 /…

Idioma del texto original: chino

2026-06-15

Mostrando 1 de 1 skills recopiladas.