Cleaner article capture: strip in-body clutter, keep section breaks and captions

[fredrivett]

What & why

Long-form articles were being captured with publisher furniture mid-body ("Read More" cards, subscribe prompts, podcast "Listen on" widgets), with section dividers dropped so sections ran together, and with photo captions rendered as ordinary paragraphs. A new pre-Readability pass (article-cleanup.ts) removes blocks whose class names match a short generic clutter list — guarded so anything holding prose, a large share of the page's text, the article itself or a tweet marker is kept — turns SVG/empty dividers and * * * paragraphs into <hr>, and turns classed caption divs into <figcaption> (now kept in the markdown). The reader gives the paragraph after each section break a drop cap (our own typography, since source drop caps are mostly CSS-only) and sets captions small, muted and tight under their image.

Checklist

• User-facing change → added/updated a PostHog event (and a funnel/insight for a new flow), or N/A — N/A, improves existing capture/reader rendering, no new action
• User-facing or service change → docs updated (README features/roadmap/services, .env.example, in-app help, CONTRIBUTING.md, onboarding checklist), or N/A — N/A, no new feature or service
• New error paths report via captureServerException / the error boundary, or N/A — N/A, runs inside the existing never-throw extraction
• New behavior is covered by tests, or N/A — unit tests for matching, guards, dividers, captions and break tidying; a synthetic sectioned-magazine fixture; and a check that the real article fixtures come through unchanged
• If this fixes a recurring defect, considered a guardrail via the learn skill, or N/A — N/A

Notes

Only new captures and re-captures pick this up; existing items keep their stored content. The fixture copies a real magazine layout's structure and class names with placeholder prose. Adds an Article/ArticleDetailView story with a Sectioned variant.

🤖 Generated with Claude Code

---

Summary by cubic

Improves long-form article capture by removing in-body publisher furniture, preserving section breaks, and keeping photo captions.

A pre-Readability cleanup pass now strips "Read More" cards, subscribe prompts, podcast widgets, and other clutter whose class names match a short generic list — guarded so prose, wrappers holding a large share of the page's text, the article itself, and tweet markers are never removed. Empty or SVG-only dividers and * * * paragraphs become <hr> section breaks, and classed caption divs become <figcaption> elements that survive into the markdown. The reader gives the paragraph after each break a drop cap and renders captions as small, muted text directly under their image.

• Guards apply to the matched element itself, so a clutter class on an articleBody wrapper, or a divider/caption class on an image, no longer slips through.
• A Storybook play test asserts the rendered structure the reader's sibling selectors rely on: dropped caps after breaks and the caption sitting tight under its image.
• Only new captures and re-captures pick this up; existing stored content is unchanged.
• Real article fixtures are verified to pass through unchanged, and a synthetic magazine fixture covers the new paths.

^{Written for commit 32f8097. Summary will update on new commits.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
abode	Ignored	Preview	Oct 5, 2026 8:38pm UTC

[cubic-dev-ai]

All reported issues were addressed across 6 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>