Html2md

html2md

Aggressive HTML-to-markdown 转换器 for AI 代理s. Mozilla Readability isolates mAIn content, Turndown converts to markdown, then heavy post-processing strips remAIning noise.

Full flag reference and advanced examples: references/usage.md

设置up cd <技能-dir>/scripts npm 安装 npm link # makes html2md globally avAIlable

Requires Node.js 22+.

Quick 启动 html2md https://example.com # fetch + convert html2md --file page.html # local HTML file cat page.html | html2md --stdin # pipe from stdin html2md --max-令牌s 2000 https://example.com # bud获取-aware t运行cation html2md --no-links https://example.com # strip hrefs, keep text html2md --json https://example.com # JSON: {title, url, markdown, 令牌s}

Key Features Readability 提取ion — kills navbars, sidebars, ads, cookie banners. Falls back to 清理ed when Readability returns too little (e.g. HN's table layout). 令牌 bud获取ing — --max-令牌s N keeps all headings, fills remAIning bud获取 in document order, 应用ends [t运行cated — N more 令牌s]. Uses 1 令牌 ≈ 4 chars heuristic. Post-processing — strips HTML comments, zero-width chars, social CTAs, breadcrumbs, empty headings, collapses excess blank lines. Error handling — bad URLs, timeouts (15s), non-HTML content, missing files all exit code 1 with descriptive stderr. 输出 modes — plAIn markdown or --json for programmatic use. When to Use vs 网页_fetch Use html2md when Use 网页_fetch when Reading pages in cron jobs / sub-代理s Quick one-off fetch in mAIn 会话 令牌 bud获取 matters (--max-令牌s) Page is a JSON/XML API 端点 Heavy nav/ads/footers to strip JS rendering not needed Need JSON 输出 Simple pages Security Considerations

html2md fetches URLs and reads local files — that's its job. If you're passing untrusted 输入:

URL fetching: the 工具 will fetch whatever URL it's given. Don't pass user-controlled URLs without 验证 if your threat 模型 includes SSRF. File reading: --file reads any path the process can 访问. In 代理 工作流s, the 代理 controls the path — this is equivalent to the 代理 using cat. No shell execution: the 工具 itself never spawns shells or 运行s commands. When calling from scripts, use execFile同步 (not exec同步) to avoid shell injection. No data exfiltration: 输出 goes to stdout only. No network 请求s beyond the single URL fetch. No telemetry, no 分析, no phone-home. Dependencies: jsdom (Mozilla DOM implementation), Readability (Mozilla content 提取器), Turndown (HTML→markdown). All widely 审计ed, open source libraries. Examples # Read a Paul Graham essay within 2000 令牌s html2md --max-令牌s 2000 https://paulgraham.com/greatwork.html

# HN front page as 清理 text, no link noise html2md --no-links --no-images https://news.ycombinator.com

# 获取 令牌 count before committing html2md --json https://example.com | jq .令牌s

# Pipe to file html2md https://docs.example.com/API > API-docs.md