web-archive

name	web-archive
description	Archive web pages, videos, and full sites for offline use. Trigger on "archive this page", "save this URL offline", "snapshot this article", "mirror this site", "download this video", "capture this thread", or any request to preserve web content for later. SKIP for unrelated uses of "archive", for example tar/zip archives, archived issues/PRs, mailing-list archives, and so on.

Web Archive

WACZ-first offline capture for the open web. Web pages, whether static, JS-rendered, logged-in, or full crawls, go through btrix and land as a single .wacz (the format published by the Internet Archive, Library of Congress, and the Webrecorder ecosystem). Video and audio go through yt-dlp. If btrix cannot script the interaction, hand off to /playwright-cli.

Decision matrix

Input	Command	Output
Web page (static, JS-rendered, logged-in)	btrix page <url>	<slug>.wacz
Web site / docs section (full crawl)	btrix site <url> -- --pageLimit N	<slug>.wacz
Video, audio, podcast feed	yt-dlp <url>	<slug>.mkv / <slug>.opus
Multi-step interactive capture browsertrix cannot script	/playwright-cli	PDF / HTML / PNG

URL-shape heuristics, plus when to escalate from a single page to a crawl, live in references/decision-tree.md.

Output convention

Default layout:

~/Documents/archives/<host>/<YYYY-MM-DD>-<slug>.<ext>

• <host>: the URL host preserved verbatim. Do not strip www. and do not alias-normalize (www.example.com stays www.example.com, x.com stays x.com, youtu.be stays youtu.be).
• <slug>: slugify applied to the page title or, when no title is available, the last URL path segment.
• <ext>: .wacz for web pages; .mkv / .opus for yt-dlp.
• Never overwrite an existing archive without an explicit "re-archive" intent from the user.
• Override: when the user explicitly names a directory ("save to this repo", "save here"), use that instead.

Logged-in profiles persist at ~/Documents/archives/<host>/profiles/profile.tar.gz and are reused across crawls.

Tools

Tool	Reference
btrix wrapper around browsertrix-crawler (WACZ for web pages, including site crawls and logged-in captures)	references/browsertrix-wacz.md
yt-dlp (video, audio, podcast feeds)	references/yt-dlp.md
/playwright-cli (narrow fallback for interactions browsertrix cannot script)	references/playwright-handoff.md

Each reference holds the canonical invocation, flag reference, and per-host recipes. Do not restate them here.

Handing off to playwright-cli

btrix handles single-page WACZ, full-site WACZ, and logged-in captures. Hand off to /playwright-cli only when the interaction itself is the thing btrix cannot script: multi-step navigation, PDF rendering of an already-interactive page, single-element screenshots, or per-element interaction before capture. Recipe: references/playwright-handoff.md.