AGENT SKILL

x-twitter

Fetch X/Twitter posts, threads, replies, profiles, and search results via the FxTwitter (FxEmbed) API and mirror domains. Use whenever an x.com or twitter.com link appears, or the user asks about X/Twitter content. Direct scraping of x.com hits login walls and blocks.

License 许可
AGPL-3.0
所属仓库
gpdir16/tabyAgent
最近更新
2026-11-04
Stars
★ 18
Forks
4

01 前言 Front Matter

name
x-twitter
description
Fetch X/Twitter posts, threads, replies, profiles, and search results via the FxTwitter (FxEmbed) API and mirror domains. Use whenever an x.com or twitter.com link appears, or the user asks about X/Twitter content. Direct scraping of x.com hits login walls and blocks.

03 SKILL.md 正文

X / Twitter access

Direct scraping of x.com / twitter.com fails (login walls, blocks). Two working paths, in order:

  1. FxTwitter REST API (preferred). Unofficial JSON API, no auth, read-only. Base: https://api.fxtwitter.com
  2. Mirror domains. For full-page rendering or when the API is down

Call endpoints with terminal_run + curl (or Python httpx/urllib).

API rules

  • GET only. No API key, no login, no cookies. Guest public data only.
  • Always send a User-Agent (e.g. -A "tabyAgent/1.0"). Empty UA gets blocked.
  • Timeout 20s. Retry only on 429 / 502 / 503, max 3 tries, backoff 1s, 2s, 4s.
  • Every response JSON has a code field. Check code === 200, not just the HTTP status. HTTP 200 with code !== 200 is a failure.
  • ~1000 req/min per IP on paper; search gets throttled much earlier.
  • Public posts only. Protected/suspended accounts and deleted posts return 401/404.
  • fxtwitter.com the website is deprecated/gone; the api.fxtwitter.com subdomain still works.

Extract the tweet id

Supported inputs: bare numeric id, or a status URL on x.com, twitter.com, fxtwitter.com, fixupx.com, twittpr.com.

(?:x\.com|twitter\.com|fxtwitter\.com|fixupx\.com|twittpr\.com)/[^/]+/status/(\d+)

Capture group 1 is the tweet_id. The handle is not needed for lookups.

Endpoints

Need Call
Single post + metrics GET /2/status/{id}
Author's thread (upward self-chain) GET /2/thread/{id}
Thread + other users' replies GET /2/conversation/{id}?ranking_mode=likes|recency
Keyword search GET /2/search?q=...&feed=latest|top|media&count=1-100
Profile GET /2/profile/{handle}, handle or id:<numeric>
User timeline GET /2/profile/{handle}/statuses?count=&cursor=
User media only GET /2/profile/{handle}/media
Quotes / reposting users GET /2/status/{id}/quotes · /2/status/{id}/reposts

Common query params: lang (translation target, e.g. ko), cursor (from prior cursor.bottom), count (1–100), about_account=1.

curl -sS -A "tabyAgent/1.0" "https://api.fxtwitter.com/2/status/1548602399862013953?lang=ko"
curl -sS -A "tabyAgent/1.0" "https://api.fxtwitter.com/2/search?q=from%3Aelonmusk&feed=latest&count=20"
curl -sS -A "tabyAgent/1.0" "https://api.fxtwitter.com/2/conversation/1548602399862013953?ranking_mode=recency"
curl -sS -A "tabyAgent/1.0" "https://api.fxtwitter.com/2/profile/elonmusk"

Success is code: 200 plus the root key (status / results / user).

Useful status fields

id, url, text, created_at, created_timestamp, likes, reposts, quotes, replies, views (may be null), lang, author.{name,screen_name,avatar_url}, replying_to, quote (or tombstone), poll, media.photos[], media.videos[], article (X Articles).

Pagination & search notes

  • If cursor.bottom is present, request the same path with ?cursor=. If it is null or empty, you are done. Keep to ~5 pages max.
  • /2/conversation page 2+ can return HTTP 404. Known bug; return what was collected and stop.
  • Search q accepts X web operators (from:, #tag, "exact phrase", min_faves:, since:). Best-effort, not guaranteed.
  • Search is recent/popular results, not a full archive. It fails more than single-post lookups: on 502 / search_unavailable say the search upstream failed, and fall back to /2/status/{id} if a URL/id is known.
  • User timeline extras: with_replies (truthy), groupthreads, since (unix ts; 204 = nothing new).
  • Default to one page (~20 items); fetch more only if the user asks.

Routing

User input Endpoint
URL or numeric id /2/status/{id}
"thread / 타래 / 스레드" /2/thread/{id}, then /2/conversation if replies needed
"replies / 답글 / 댓글" /2/conversation/{id} (ranking_mode=recency for chronological)
"search / 찾아 / 최근글" + keyword /2/search
@handle or "profile / 프로필" /2/profile/{handle}
"their posts / 그 계정 글" /2/profile/{handle}/statuses

Mirror domains (fallback / page view)

When the API can't help (need the rendered page, media embed, or API down), rewrite the host instead of browser-automating x.com:

  • Replace x.com with fixupx.com (or xfixup.com)
  • Replace twitter.com with twittpr.com
  • Further fallbacks: fixvx.com, yt-dlp, or the browser tools on a mirror page. Try the next when one 404s/blocks.

Don'ts

  • Never mix official api.x.com schemas with this API.
  • Never mix the legacy /{handle}/status/{id} form (root key tweet) with v2 /2/... (root key status). Use v2 only.
  • No bulk search crawling; no cookies or account credentials. This API only serves guest-public data.
  • Don't report success on code !== 200 even when HTTP is 200.

Error mapping

code/HTTP meaning
400 bad request
401 private or restricted
404 not found
429 rate limited
500/502/503 upstream error

Limits: relay honestly to the user

  • Unofficial proxy; can break or shut down without notice.
  • Search is not a full archive; replies may be partial.
  • Protected accounts and deleted posts are unreachable.

04 同仓库更多 Skills