菜单
Extract distilled page content as Markdown using Chromium's on-device ML page content annotation.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | distill-page |
| description | Extract distilled page content as Markdown using Chromium's on-device ML page content annotation. |
This skill extracts high-quality, noise-free Markdown content from a webpage using Chromium's experimental, on-device Machine Learning model annotations (Page.getAnnotatedPageContent).
Unlike generic HTML-to-Markdown parsers (e.g. Turndown) which rely on markup syntax heuristics, this approach uses Chromium's visual ML layout analysis. It automatically identifies semantic roles (such as ARTICLE and MAIN) and excludes sidebars, navigation elements, footers, and popups.
node agents/skills/distill-page/scripts/distill-page.ts [--markdown] <URL>
--markdown: Output as distilled Markdown (recommended). Without this, it outputs the raw decoded protobuf JSON.--base64-stdin: Instead of fetching via Playwright, decode a base64-encoded protobuf payload passed via stdin.# Get markdown representation of page
node agents/skills/distill-page/scripts/distill-page.ts --markdown https://example.com
# Get raw JSON structure of annotated layout tree
node agents/skills/distill-page/scripts/distill-page.ts https://example.com
headless: false) and requires Chromium feature flags:
--enable-dom-distiller--enable-features=OptimizationHints,PageContentAnnotation,OptimizationGuideModelDownloadingFetch a URL using Playwright and convert its content to Markdown.
Run a Google Search and capture the AI response/Overview.
Gerrit helpers for fetching unresolved comments and CL diffs.
Set up or debug npm Trusted Publishing (OIDC) from GitHub Actions. Handles permissions, metadata validation, and provenance.
Paul's modern stack conventions for new projects (pnpm, native node test, esbuild, buildless-types). Consult when starting a new project or repository.
How to effectively use QMD (Quick Markdown Search) to pull in extra relevant context, run multi-query searches, extract high-scoring chunk IDs, and fetch their contents. Use this skill when asked to search across a large markdown knowledge base, run QMD queries, or retrieve deep context using the qmd CLI.