hawkongz

High-accuracy PDF content extraction using MinerU (Shanghai AI Lab). Use this whenever the user needs to extract text, formulas, tables, or images from a complex PDF — especially academic papers, multi-column layouts, scanned documents, or any PDF where pypdf produces garbled /Cxx formula output. Trigger on: "MinerU", "解析这篇论文", "提取PDF公式", "这个PDF乱码了", "扫描件OCR", "双栏PDF", or any request involving PDF formula/image extraction. Even if the user doesn't say "MinerU" explicitly, suggest it when they complain about garbled PDF text or need high-quality extraction. For simple single-column text-only PDFs, the default pdf skill is faster.

使用 MinerU（上海 AI Lab）进行高精度 PDF 内容提取。 当用户需要从复杂 PDF 中提取文本、公式、表格或图片时使用此 skill — 尤其是学术论文、多栏排版、扫描文档，或任何 pypdf 产生 /Cxx 乱码的 PDF。 触发词："MinerU"、"解析这篇论文"、"提取 PDF 公式"、"这个 PDF 乱码了"、 "扫描件 OCR"、"双栏 PDF"，或任何涉及 PDF 公式 / 图片提取的请求。 即使用户没有明确说 "MinerU"，当他们抱怨 PDF 文本乱码或需要高质量提取时， 主动建议使用此 skill。对于简单的单栏纯文字 PDF，默认 pdf skill 更快。

Trigger when users say "explain with a story," "tell me a fable," "give me an analogy," "ELI5," or "help me understand intuitively" about a concept. Uses fables to illuminate abstract concepts — the story never mentions the term, only revealing it at the end.

当用户说「用故事解释」「讲个寓言」「打个比方」「通俗理解」某个概念时触发。用寓言故事诠释抽象概念，故事中不出现学名，结尾才揭晓。

Use when the user wants to share a project to GitHub. Runs a full pipeline: root audit, file naming and structure, document fixes (headings, code-block language tags, table alignment, explicit command interpreters), generate missing files (LICENSE, .gitignore, CONTRIBUTING.md, Issue/PR templates), README rewrite (centered title, TOC), bilingual setup, git init, conventional commit, gh repo create, push, set repo About (description + topics), semantic tag. Triggers on: 'share to GitHub', '/GitHub格式规范', '上传到 GitHub', '发布到 GitHub', '推到 GitHub', '发到 GitHub', '上传GitHub', '发布GitHub'.

当用户想把项目分享到 GitHub 时使用。执行完整流水线：根目录审查、文件命名与目录结构、文档修复（标题层级、代码块语言标注、表格对齐、命令显式解释器）、生成缺失文件（LICENSE .gitignore CONTRIBUTING.md Issue/PR 模板）、README 重写（居中标题 目录）、双语设置、git init、约定式提交、gh repo create、push、设置仓库 About（描述 话题标签）、语义化标签。触发词：分享到 GitHub、/GitHub格式规范、上传到 GitHub、发布到 GitHub、推到 GitHub、发到 GitHub、上传GitHub、发布GitHub。

Use when the user shares an image that Claude can't natively see, or when the user asks to analyze/describe/understand an image. Route every image through this skill — send the image path to eyes.py, which calls a vision model (default qwen3-vl-plus) via 百炼 DashScope and returns a Chinese text description. Then answer the user's question based on that description.