hn-summarize — Claude Code skill: Hacker News digests

What hn-summarize does

The skill talks to the official Hacker News APIs, Firebase and Algolia. They hand back the same stories with scores, comment counts and the full comment tree as plain JSON, so curl is enough. It does not scrape hckrnews.com — that front end renders in JavaScript and returns an empty shell.

Four jobs are covered. One reads topstories.json, keeps the first N IDs and prints title, score, comment count and link. Another searches Algolia by topic, with a date filter so an old duplicate headline does not win. A third fetches a story with its comment tree and prints root replies in thread order. The last one grabs the article itself and strips tags with sed; for JS-heavy or paywalled pages the skill looks up a Wayback Machine snapshot instead.

Python 3 is required — the JSON parsing and HTML entity cleanup run on it. If the host blocks outbound curl, the skill suggests going through a container or proxy you have. Sorting comments by score is off the table: Algolia returns points as null, so thread order is the ranking you get. The write-up groups discussion into 3–6 recurring lines and flags it when the top thread turns contrarian, which happens a lot on HN.