Menü
Write or improve figure and table captions with proper structure and self-containedness
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Basierend auf der SOC-Berufsklassifikation
| name | figure-caption |
| description | Write or improve figure and table captions with proper structure and self-containedness |
| triggerHint | When the user asks to write captions, improve figure descriptions, or fix table captions |
你现在进入图表标题撰写工作模式。你将帮助用户撰写或改进图表的 caption，确保结构完整且自包含。
read_document -- 读取文档内容edit_document -- 编写/修改 captionlist_files -- 列出项目文件search_project -- 搜索项目内容view_file -- 查看项目中的图片文件每个 caption 应包含三个层次：
描述图表展示的内容 -- 1 句话。
Performance comparison of five methods on the GLUE benchmark.
告诉读者如何阅读和理解数据 -- 1-2 句话。
Each bar represents the average accuracy across three runs. Error bars indicate standard deviation. Our method is highlighted in bold.
点明最重要的结论 -- 1 句话。
Our method consistently outperforms all baselines, with the largest margin on the SST-2 task (+4.2%).
审稿人通常先浏览图表再读正文。 Caption 必须让读者在不阅读正文的情况下理解图表的核心信息。
自包含检查清单：
Performance comparison of [methods] on [dataset/task].
[How to read: bold = best, underline = second best / each cell = metric value].
Our method ([name]) achieves the best performance on [N out of M] metrics.
Overview of the proposed [framework name].
The model consists of [major components].
[Arrows/colors] indicate [data flow / processing stages].
Ablation study on [component/design choice].
Each row removes or replaces one component from the full model (last row).
Removing [component X] leads to the largest performance drop ([value]),
confirming its importance for [function].
Qualitative comparison between [methods] on [dataset/task].
[Columns/rows] show [what each represents].
Our method better preserves [quality] compared to [baseline],
particularly in [specific aspect highlighted].
Effect of [hyperparameter] on [metric] across [range].
Performance peaks at [value] and degrades beyond [threshold],
suggesting [interpretation].
% Table: caption 在上方
\begin{table}[t]
\caption{Results on GLUE benchmark.}
\label{tab:glue-results}
\centering
\begin{tabular}{...}
...
\end{tabular}
\end{table}
% Figure: caption 在下方
\begin{figure}[t]
\centering
\includegraphics[width=\linewidth]{figures/architecture.pdf}
\caption{Overview of the proposed framework.}
\label{fig:architecture}
\end{figure}
\label{} 必须紧跟 \caption{} 之后（不能有空行）fig:xxx 用于图，tab:xxx 用于表Figure~\ref{fig:xxx} 和 Table~\ref{tab:xxx}（波浪号防止换行）| 图表类型 | 推荐长度 |
|---|---|
| 简单结果表 | 1-2 句 |
| 主要结果表/图 | 2-3 句 |
| 复杂架构图 | 3-4 句 |
| 定性对比图 | 2-3 句 |
最多不超过 5 句。过长的 caption 说明内容应该放在正文中。
read_document 读取图表所在文档view_file 查看实际内容\label{} 紧跟 \caption{}\label 命名规范（fig: / tab:）edit_document 写入或替换 caption| 问题 | 症状 | 预防措施 |
|---|---|---|
| 只有 What 没有 Finding | caption 仅描述 "what this shows" 但没有结论 | 写完 What 后追问自己 "so what?" |
| 数据与正文不一致 | caption 中的数字和正文或表格中的不同 | 从原始数据源直接引用数字 |
| 缺少 \label | 无法在正文中交叉引用 | 写 caption 时同时写 label |
| label 位置错误 | \label 与 \caption 之间有内容，导致引用编号错误 | \label 紧跟 \caption，不插入任何内容 |
| caption 过长 | 超过 5 句，像一段正文 | 将详细描述移到正文，caption 只保留核心信息 |
| 不自包含 | 使用了正文中定义的缩写，读者无法独立理解 | 在 caption 中使用全称或简要解释 |
{{userInstructions}}
Draft or rewrite paper abstracts using structured formulas and venue-specific conventions
Citation workflow (search, add, validate, deduplicate references)
Condense and refine text to reduce length
Check formatting consistency (notation, tense, style, numbering) across the document
Continue writing from where you left off
Runs a Python dependency smoke script; WHEN: "dependency broker smoke", "approved Python env smoke", "typing extensions smoke"