Reformats CourtListener opinions for better read-later parsing.

CourtListener serves opinion text wrapped in a single <pre class="inline"> block, with hard line-wraps and page-break characters baked into the source. On courtlistener.com this is styled to look normal, but read-later tools like Readwise Reader and GoodLinks strip site CSS and fall back to default <pre> rendering — a fixed-width, non-wrapping code block.

Given a CourtListener opinion URL, this val:

1. Extracts the cluster ID from the URL.
2. Looks up the cluster and its lead opinion via the CourtListener REST v4 API.
3. Reflows the opinion's html_with_citations field: strips page-break artifacts, joins hard-wrapped lines back into paragraphs, and re-wraps everything in <p> tags.
4. Returns a clean, minimally-styled HTML page that read-later tools parse normally.

1. Get a free CourtListener API token from your CourtListener profile.
2. In this val's settings, add an environment variable/secret named COURTLISTENER_TOKEN with that value.
3. Save. Val Town deploys automatically — no separate deploy step.

https://joeross-courtlistenerreflow.web.val.run/?url=<courtlistener-opinion-url>

url accepts either a full opinion URL or a bare cluster ID.

From any CourtListener opinion page, run:

javascript:location.href='https://joeross-courtlistenerreflow.web.val.run/?url='+encodeURIComponent(location.href)

This redirects to the cleaned version, which can then be saved to Readwise Reader or GoodLinks as usual.

• Only the lead opinion (sub_opinions[0]) is fetched. Concurrences and dissents are not included.
• Page-break stripping assumes a docket-number-then-page-number pattern (e.g. A-2182-24 followed by a page number before a form feed). Courts with different pagination formatting may leak stray page numbers into the text.
• CourtListener's free-tier API rate limit is 5 requests/minute, 50/hour, 125/day. Fine for one-off bookmarklet use; a constraint if wired into bulk/RSS processing.

Built with Claude.