X Public Reader skill

x.com is login-walled, so agents that browse or scrape it get a teaser, a login page or a 403. This skill gives any AGNT agent a reliable way to read public X data instead: profiles, full timelines with paging, single posts, threads, search results, photos and videos (with the highest-bitrate MP4), quoted posts, community notes, and long-form X Articles flattened to markdown.

It uses FxTwitter's public JSON API as the primary source and Twitter's own embedded-timeline endpoint as a fallback, and it encodes the failure modes we measured while building it. FxTwitter returns 404 at random for real accounts (4 of 12 timeline requests), so one 404 is never reported as "user not found". Timeline pages overlap, so results are de-duplicated. The fallback can return HTTP 200 with an empty timeline, which is treated as an error rather than "no posts". Rate limits are honoured from the server's own reset headers.

Output is normalised JSON that tells the agent what it is looking at: whether a post is a repost, a reply or a quote, and whether the full text lives in an X Article. That way it credits the right author and fetches context before summarising.

Includes a zero-dependency Node script (profile, posts, post, thread, search) with 16 tests, and an inline recipe for runtimes that cannot run scripts. Public data only and read-only: no protected accounts, DMs or posting.

A skill is a SKILL.md an agent reads when a task matches it. Once installed, any AGNT agent can use it, with any model provider.