All to Markdown

将任意文件（PDF、Word、Excel、PPT、图片、音频、网页等）转换为 Markdown

Security Scan

The skill is coherent with its stated purpose (it wraps the markitdown CLI to convert files/URLs to Markdown), requests no unrelated credentials, and contains no code that exfiltrates data — the only network activity is fetching user-supplied URLs and optional calls to OpenAI when the user provides an API key.

Purpose & Capability

The skill's name/description, SKILL.md, config.json, and run.sh consistently describe a wrapper around the markitdown CLI to convert files/URLs to Markdown. Minor inconsistency: top-level registry metadata listed no required binaries, while config.json and scripts/run.sh require the markitdown CLI (installed via pip). This is explainable (packaging metadata omission) but worth noting.

Instruction Scope

SKILL.md instructs the agent to run scripts/run.sh with a user-provided path or URL and to pass through options to markitdown; the runtime script only checks for markitdown and execs it. Instructions limit network access to fetching provided URLs/YouTube and clearly mark OPENAI_API_KEY as optional for AI-enhanced image/OCR features.

Install Mechanism

No automatic install spec in the skill bundle; SKILL.md and config.json recommend installing markitdown via pip (pip install 'markitdown[all]'). This is a standard public-package install path and the bundle does not download arbitrary archives or run remote installers.

Credentials

No required environment variables. OPENAI_API_KEY is declared optional and used only for AI/image/OCR enhancements as documented — this is proportional and expected. The skill does not request unrelated credentials or access.

Persistence & Privilege

The skill does not request always:true, does not modify other skills or system-wide config, and is user-invocable only. It executes markitdown at runtime and does not persist elevated privileges.

Assessment

This skill is a thin wrapper around the community markitdown tool and appears internally consistent. Before installing: (1) review and install markitdown from PyPI in a virtualenv (pip install 'markitdown[all]') rather than running untrusted installers; (2) be aware that converting a URL causes the tool to fetch remote content and enabling OPENAI_API_KEY will send image/OCR content to OpenAI — do not pass sensitive documents or secrets if you enable that feature; (3) consider inspecting markitdown's code or its installed plugins if you need higher assurance; and (4) note the small metadata mismatch about the required binary (markitdown) — ensure that CLI is present on the system before invoking the skill.

基于 Microsoft MarkItDown，将任意格式的文件或 URL 转换为 Markdown，便于 LLM 分析和处理。

支持格式

类型	格式
文档	PDF、DOCX、PPTX、XLSX、XLS、EPUB、MSG
数据	CSV、JSON、XML
图片	JPG、PNG 等（含 EXIF 元数据，可选 OCR）
音频	WAV、MP3（含语音转录，需 OpenAI Key）
网页	HTML、YouTube URL（含字幕提取）
压缩	ZIP（逐文件转换）

前置要求

安装 markitdown：

pip install 'markitdown[all]'

给 AI 的使用说明

当用户需要将文件或 URL 转换为 Markdown 时，使用以下命令：

scripts/run.sh <文件路径或URL>

可选标志：

• -o <输出文件> — 保存到文件
• --use-plugins — 启用插件（如 markitdown-ocr）

重要原则：

• 转换结果直接输出到 stdout，可供 AI 直接读取分析
• 文件路径使用用户提供的实际路径，不要假设
• 转换大型文件时提前告知用户可能需要较长时间

使用示例

示例 1：转换 PDF

用户：帮我把这个 PDF 转成 Markdown，以便我分析内容

AI 执行：

scripts/run.sh /path/to/document.pdf

示例 2：转换并保存

用户：把这个 Excel 转成 Markdown 文件保存

AI 执行：

scripts/run.sh /path/to/data.xlsx -o output.md

示例 3：转换网页

用户：把这篇文章转成 Markdown

AI 执行：

scripts/run.sh https://example.com/article.html

示例 4：提取 YouTube 字幕

用户：把这个 YouTube 视频的内容提取出来

AI 执行：

scripts/run.sh https://www.youtube.com/watch?v=xxx

可选 AI 增强功能

设置 OPENAI_API_KEY 后，markitdown 可对图片生成 AI 描述：

OPENAI_API_KEY=sk-xxx scripts/run.sh image.jpg

安全说明

• 仅在本地执行文件转换，不发送文件内容到远程服务
• 转换 URL 时会访问对应网络地址
• 启用 LLM 功能时，图片内容会发送到 OpenAI API