hackernews-frontpage

Run the skill via `$B skill run hackernews-frontpage`. It navigates to news.ycombinator.com through the browser daemon, parses each story row, and prints a single JSON document with stories and a count field to stdout.

Each story includes a 1-based rank, HN item id, title, outbound URL, points, and comment count. Job postings that lack scores return null for points and comments, and 'discuss' links count as zero comments.

No Hacker News credentials are needed since the front page is public. The script does require the gstack browser daemon, authenticated via GSTACK_PORT and GSTACK_SKILL_TOKEN env vars set by the harness, or a .gstack/browse.json state file when run directly.

Scrapers break when HN rotates its HTML structure and selectors no longer match. This skill mitigates that with a bundled HTML fixture test: the parser is a pure function tested against captured markup, so breakage fails the test before users notice.

Yes. The parseStoriesFromHtml function is exported separately from the daemon entry point, so script.test.ts exercises it directly against the bundled fixture using bun:test with no network or daemon required.