Skip to content

/writing-polish

PaperFit writing polish expertise for LaTeX paper layout.

From plugin
paperfit
3379 skills6 agents9 commands

Claude Code

$ npx -y skills add OpenRaiser/PaperFit --skill writing-polish --agent claude-code

SKILL.md

writing-polish.SKILL.md
name: writing-polish
description: "PaperFit writing polish expertise for LaTeX paper layout."

Writing Polish Skill

概述

本技能为 **Semantic Polish Agent** 提供具体、可执行的语义微调策略与禁区规则。它定义了在排版手段用尽后,如何通过最小化文字增删来消除孤行寡行、控制页数预算或优化末页留白,同时严格保持学术内容的原意、数据和结论不变。

该技能不直接被 `orchestrator-agent` 调用,而是作为 `semantic-polish-agent` 的知识库和行为规范。所有语义级改写必须遵循本技能中定义的技巧和约束。

Source-changing 边界:语义级改写是高风险 source mutation。默认只能生成候选改写、风险和 approval 状态;只有显式 `--apply` 或宿主等价授权存在时才允许写回。多轮语义改写必须同时显式启用 `--apply --max-rounds N`,并且每轮都通过 approval carry-forward、artifact freshness、candidate approval scope gate 和 gatekeeper `CONTINUE`;超出既有 approval scope 的语义修改必须重新获批。

适用场景

| 触发缺陷 | 操作方向 | 允许的改写幅度 | |----------|----------|---------------| | A1(孤行寡行) | 缩短 1-2 行 | 增删 3-8 词 | | A2(末页留白) | 扩展 2-4 行 | 增加 20-50 词 | | A3(页数预算) | 缩短或扩展多行 | 按需,但需分段执行 | | 用户主动请求 | 精炼或扩写特定段落 | 用户指定 |

输入规范

| 输入项 | 来源 | 说明 | |--------|------|------| | 目标段落源码 | `semantic-polish-agent` 提取 | 需改写的一个或多个完整段落 | | 改写目标 | 调用方请求 | `shorten` 或 `expand`,及期望行数变化 | | 写作规范 | `config/writing_rules.yaml` | 时态、术语、禁用词等约束 | | 上下文段落 | `semantic-polish-agent` 提取 | 前后各一段,用于保证语义连贯 |

输出规范

本技能输出改写后的文本及变更元数据,供 `semantic-polish-agent` 整合为最终报告。

{
  "skill": "writing-polish",
  "changes": [
    {
      "paragraph_id": 3,
      "action": "shorten",
      "net_word_change": -6,
      "before_snippet": "It is worth noting that our method achieves state-of-the-art performance on several benchmark datasets.",
      "after_snippet": "Our method achieves state-of-the-art results on several benchmarks.",
      "techniques_used": ["remove_redundant", "phrase_to_word"],
      "rationale": "移除冗余修饰词,将 'achieves state-of-the-art performance on several benchmark datasets' 压缩为 'achieves state-of-the-art results on several benchmarks'。语义等价,数据未变。"
    }
  ],
  "warnings": []
}

改写策略

通用原则

1. **最小修改优先**:能改一词不改一句,能改一句不改一段。 2. **保持学术严谨**:绝不改变数据值、引用标记、专有名词、核心声明。 3. **局部影响评估**:每次改写后需编译验证,确保不引入新的孤行或溢出。 4. **可逆性**:保留改写前文本,便于人工审查或回滚。

---

策略组 1:缩短(Shorten)

目标:在不损失信息的前提下减少字数/行数。

技巧 1.1:删除冗余修饰词

移除对学术内容无实质贡献的修饰语。

  • 删除强调性副词:`very`、`quite`、`extremely`、`highly`
  • 删除填充短语:`It is worth noting that`、`It should be emphasized that`、`It is important to mention that`
  • 删除冗余限定:`in a certain sense`、`to some extent`

示例:

修改前:It is worth noting that our method achieves very competitive performance.
修改后:Our method achieves competitive performance.
减少:5 词

技巧 1.2:短语替换为单词

用更简洁的单词或缩写替代多词短语。

| 原短语 | 替换为 | |--------|--------| | `in order to` | `to` | | `a large number of` | `many` | | `due to the fact that` | `because` | | `at the present time` | `now` | | `state-of-the-art methods` | `SOTA methods`(需已定义) | | `with respect to` | `regarding` 或 `on` |

示例:

修改前:We conduct experiments in order to evaluate the performance of the proposed approach.
修改后:We conduct experiments to evaluate our approach.
减少:5 词

技巧 1.3:被动语态转主动语态

主动语态通常更简短且更有力。

修改前:The experiments were conducted by us on three datasets.
修改后:We conducted experiments on three datasets.
减少:3 词

技巧 1.4:合并相邻短句

将两个紧密相关的短句合并为一句。

修改前:We used the Adam optimizer. The learning rate was set to 1e-4.
修改后:We used Adam with a learning rate of 1e-4.
减少:6 词

技巧 1.5:使用标准学术缩写

在全文首次定义后,使用公认缩写。

| 原词 | 缩写 | |------|------| | `state-of-the-art` | `SOTA` | | `natural language processing` | `NLP` | | `mean average precision` | `mAP` |

修改前:Our method outperforms previous state-of-the-art approaches on the natural language processing benchmark.
修改后:Our method outperforms previous SOTA approaches on the NLP benchmark.
减少:5 词(假设 SOTA/NLP 已定义)

技巧 1.6:简化从句结构

将定语从句压缩为分词短语或前置定语。

修改前:The model which is trained on ImageNet achieves high accuracy.
修改后:The ImageNet-trained model achieves high accuracy.
减少:3 词

---

策略组 2:扩展(Expand)

目标:在不注水的前提下增加有实质内容的文字。

技巧 2.1:显式化隐含因果关系

在结果陈述后补充简短的原因解释。

修改前:Our method outperforms the baseline by 3.2%.
修改后:Our method outperforms the baseline by 3.2%, likely because the attention mechanism better captures long-range dependencies.
增加:11 词

技巧 2.2:补充结果解释

在表格或数据引用后,增加一句对关键发现的解读。

修改前:Table 2 shows the ablation results.
修改后:Table 2 summarizes the ablation study. Removing the temporal module causes a significant drop of 5.1%, confirming its importance for sequential modeling.
增加:18 词

技巧 2.3:强化与相关工作的对比

在提及已有工作时,增加具体的差异说明。

修改前:Unlike previous work, we use a transformer-based architecture.
修改后:Unlike previous work that relied on recurrent networks with limited parallelization, we adopt a transformer architecture that scales more efficiently to long sequences.
增加:14 词

技巧 2.4:添加局限性讨论

在结论或讨论部分,补充一句对当前方法局限性的客观陈述。

修改前:Future work will explore larger-scale datasets.
修改后:Future work will explore larger-scale datasets. A current limitation is the reliance on pre-trained word embeddings, which may not fully capture domain-specific terminology.
增加:17 词

技巧 2.5:拆分长句为短句

通过增加句号拆分长句,可在不显著增加内容的情况下扩展行数。

修改前:Our method consists of three components: an encoder, a decoder, and a refinement module.
修改后:Our method consists of three components. First, the encoder extracts features from the input. Second, the decoder generates initial predictions. Finally, the refinement module iteratively improves the output.
增加:14 词,行数增加更多

技巧 2.6:补充技术细节(谨慎)

在不泄露未公开信息的前提下,可适当补充已在论文其他部分出现过的技术细节。

修改前:We use a standard cross-entropy loss.
修改后:We use a standard cross-entropy loss with label smoothing of 0.1, following common practice in image classification.
增加:10 词

---

改写禁区(绝对禁止)

以下操作 **严禁** 进行,违反任何一条都不得输出修改。

禁区 1:篡改数据与结果

  • 不得修改任何数值、百分比、指标名称。
  • 不得增删或更改实验设定、数据集名称、模型参数。
  • 不得修改表格中的任何单元格内容。

禁区 2:编造内容

  • 不得引入原论文中不存在的引用、相关工作、方法细节。
  • 不得虚构实验、消融研究、用户调查。
  • 不得添加未经作者确认的局限性或未来工作方向(除非是论文其他部分明确提及的内容)。

禁区 3:改变核心声明

  • 不得弱化或夸大论文的贡献声明。
  • 不得修改结论段落中的主要论断。
  • 不得改变任何 `\ref{}` 或 `\cite{}` 的引用关系。

禁区 4:引入非学术表达

  • 不得使用口语化、情绪化、主观化的语言。
  • 不得添加无意义的填充句(如 `This is a very interesting result.` 后无任何分析)。
  • 不得违反 `con
Read more
@openraiser
Ships withpaperfit

Vision-in-the-Loop Academic Typesetting Agent System PaperFit 是面向 Claude Code、Codex、Cursor 的视觉闭环 LaTeX 排版系统。PaperFit 组织“渲染页图—视觉诊断—受约束源码修复—重新渲染与视觉复验”的结构化循环,由当前多模态宿主执行。实际页图反馈是诊断和每个修复候选验收的必要环节,每个完整候选重新渲染后都检查全部物理页面,再决定保留、修改或回滚;独立视觉 API

Stars on GitHub

Momentum

stars in total
337
Star history of OpenRaiser/PaperFit on GitHub, a chart by star-history.com

Other skills on paperfit.

paperfit
Auto-invoked

paperfit

Diagnose and repair LaTeX paper layout through PaperFit's structured visual closed loop over rendered pages, source-grounded…

@openraiser@openraiserInstall Skill
consistency-polisher
Auto-invoked

consistency-polisher

Inspect and repair local table, figure, and caption presentation while preserving content and venue formatting.

@openraiser@openraiserInstall Skill
float-optimizer
Auto-invoked

float-optimizer

PaperFit advice for LaTeX figure/table placement, width, clustering and breakable-table flow, verified against current page…

@openraiser@openraiserInstall Skill

More skills from other plugins.

prompt-lookup

prompt-lookup

Activates when the user asks about AI prompts, needs prompt templates, wants to search for prompts, or mentions prompts.chat. Use…

@gh-f@gh-fInstall Skill
ppt-master

ppt-master

AI-driven presentation workflow for generating editable PPTX decks and slides, reconstructing page visuals, creating reusable…

@hugohe3@hugohe3Install Skill
morph-ppt

morph-ppt

Use this skill when the user wants a .pptx with smooth cross-slide animation — PowerPoint Morph transitions, Keynote-style…

@iofficeai@iofficeaiInstall Skill

Questions about writing-polish.

Which coding agents can use writing-polish?
writing-polish installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf with the skills CLI, using each agent's flag shown on this page. Flowy's auto-invocation, which fires a skill when your prompt matches, runs in Claude Code only.