channel-hubspot-video

name	channel-hubspot-video
description	HubSpot CMS pages whose content is a lazy-loaded HubSpot Video (Mux underneath) rather than text — a static fetch finds no player at all.
argument-hint	--stage harvest --capture-dir <dir> [--url <url>]
user-invocable	false

Channel: HubSpot Video

You harvest a site built on HubSpot CMS whose pages carry HubSpot Video players. This file is authoritative for how that platform is captured. The job already carries the resolved config (scope, assets, transcript policy) — honor it; never re-ask the operator.

This unit is platform-general, not site-specific. Every HubSpot CMS site runs on its own domain and themes its own markup, so match.hosts is empty and custom_domains: true — skills find reports this unit under possible[] for an unmatched host, and the Fingerprints section below is how you settle whether it applies. Once installed, the copy is wiki-owned: put the site's own selectors, URL map and exclusions in it, and skills list reporting it as diverged is provenance, not a problem.

Stages

• --stage harvest — a claimed job carries skill: channel-hubspot-video. Capture the URL into --capture-dir with scripts/capture_hubspot_video.py and the platform knowledge below.
• --stage process — not declared. The generic scaffolder builds the note; give it your site's extract block (below) so it strips the theme's chrome.

Fingerprints — is this venue HubSpot CMS?

possible[] sent you here on a host nobody enumerated. Confirm before installing:

• Assets served under /hubfs/ or /hs-fs/hubfs/.
• hubspot in the page scripts; _hcms/ paths disallowed in robots.txt.
• Internal nav links carrying ?hsLang=<lang>.
• A player iframe at play.hubspotvideo.com/v/<portal>/id/<video>.

Any two of those together is conclusive. The last one alone means this unit's Media section applies even if the rest of the site is not HubSpot-themed.

Discovery

<host>/sitemap.xml is typically one flat urlset covering the whole site — the enumeration route worth using. Page sidebars are frequently partial or carry stale spellings, so prefer the sitemap and let the sidebar be a cross-check.

Strip ?hsLang= from discovered URLs before queueing, or the same page is queued twice under two URLs. The wiki's watch should also carry an --exclude-url '*hsLang=*'.

Media — the reason this unit exists

1. The iframe is lazy. The markup carries data-hsv-src, never src. A curl or Firecrawl fetch therefore finds no player at all — asset detection returns images only, and the capture looks complete while missing its entire content. Rendering is mandatory, not an optimization.
2. HubSpot Video is Mux underneath. After render, the network log holds image.mux.com/<PLAYBACK_ID>/storyboard.vtt — that request fires even before play is clicked, and is the most reliable place to read the id.
3. The manifests the player fetches are signed and expiring. manifest-*.edgemv.mux.com/…/rendition.m3u8 carry expires= and are per-rendition. Do not download those — they die in hours. Rewrite to the stable master https://stream.mux.com/<PLAYBACK_ID>.m3u8, which does not expire and offers every rendition to yt-dlp.
4. No DRM, and typically no <track> captions. captions/ stays empty and the generated transcript is the page's only text — which is why the watch runs transcript: always and why assets: download-audio is worth considering when only the transcript is wanted.
5. Drop verifi.podscribe.com/tag from the asset manifest: an analytics beacon the generic detector types as an image.

A page with no player at all is a real shape, not a failure: a pointer page whose payload is an external link (a podcast host, a PDF, a YouTube mirror). Record the external URL in capture.json and complete the job rather than flagging it forever.

Content extraction — set this per site

HubSpot themes vary completely between customers, so this unit ships no selectors. After the first capture, put the site's own rules in this installed copy's manifest.json, where both pipeline stages read them:

"extract": {
  "content_selector": "main#main-content",
  "drop_selectors": [".cta-block", ".legal-disclaimer", ".course-modules"],
  "title_selector": "main#main-content h2"
}

• content_selector — the page's real content root.
• drop_selectors — theme chrome that would otherwise be the bulk of every note: repeated module navs, CTA blocks, lead forms, legal disclaimers. These are removed at the DOM level, before markdown conversion, which is the only point where they are still distinguishable from prose.
• title_selector — set it whenever the site reuses one <title> across a section, which HubSpot sites commonly do. Without it every page in a course is named identically, and title is an indexed field.

A thin page.md is expected here and is not a truncated capture — the content genuinely is the video. Do not record a partial verdict for a short body on this platform.

Scripts

Resolve relative to this file; run from the wiki root.

• scripts/capture_hubspot_video.py render <url> --capture-dir <dir> — renders the page, clicks play, writes page.html, net.json, and meta.json (title, Mux playback id, stable stream URL). Exit 4 = rendered but no video resolved, which is the pointer-page case above.
• scripts/capture_hubspot_video.py patch-assets <assets.json> --meta <meta.json> — applies the Media rules to a manifest from the plugin's assets.py detect: drops the expiring edgemv manifests and the podscribe beacon, appends the stable Mux master.

Budgeting

Video-first pages are much longer than they look. One measured corpus averaged ~28 minutes per page across 75 videos — 31.7 GB and ~35 hours of audio at yt-dlp's default format pick. Estimate before committing to a section, and prefer assets: download-audio when the transcript is the only thing wanted: the same corpus would have been roughly 2 GB.

Cap concurrency at 2 workers with 2–5 s jitter, per the fan-out defaults in llm-wiki-ops reference pipeline (domain_limit, request_delay_s).

Quirks log

• 2026-07-31 — Sites in this family have shipped misspelled sidebar links that 301 to the sitemap spelling. Following them works, but enumerating from the sitemap keeps capture directories and note names correctly spelled.