x-post-reader

name	x-post-reader
description	Reads public X (Twitter) posts without auth — single tweets, a user's timeline, top ~100 tweets, or topical tweets. Use when the user pastes an x.com/twitter.com status URL or asks for a user's tweets. For authenticated X API operations (mentions, followers, search), use the x-api skill, if installed (needs X_BEARER_TOKEN). Triggers: tweet URL, x.com, what has @handle posted, top tweets.

X Post Reader: single posts, recent timeline, top tweets, and topical search — no auth

When to use

Use this skill any time you encounter a public X/Twitter status URL, e.g.:

• https://x.com/{handle}/status/{tweet_id}
• https://twitter.com/{handle}/status/{tweet_id}
• https://x.com/{handle}/status/{tweet_id}?s=20 (query params are fine — strip them)

Triggers include: "read this tweet", "what does this post say", "summarize this thread", "draft a reply to this", "quote tweet this", or any pasted status URL.

Do NOT scrape x.com directly — it requires JS and login walls. Use VxTwitter's JSON endpoint instead.

How to fetch

Swap the host for api.vxtwitter.com and keep the /{handle}/status/{tweet_id} path:

curl -sS "https://api.vxtwitter.com/{handle}/status/{tweet_id}"

No auth header. No bearer token. Works for any public post.

If VxTwitter errors or times out, swap the host to api.fxtwitter.com (same path shape, slightly different field names) before giving up.

Example

curl -sS "https://api.vxtwitter.com/jack/status/20"

Returns JSON like:

{
  "text": "just setting up my twttr",
  "user_name": "jack",
  "user_screen_name": "jack",
  "date": "Tue Mar 21 20:50:14 +0000 2006",
  "likes": 309499,
  "retweets": 126335,
  "replies": 17881,
  "hasMedia": false,
  "mediaURLs": [],
  "media_extended": [],
  "qrt": null,
  "replyingTo": null,
  "communityNote": null,
  "lang": "en",
  "tweetURL": "https://twitter.com/jack/status/20"
}

Key fields

Field	What it gives you
text	The post body. This is what you usually want.
user_name / user_screen_name	Display name / @handle
date	Human-readable timestamp
likes / retweets / replies	Public metrics
hasMedia / mediaURLs / media_extended	Image/video URLs if attached
qrt	If this post quote-tweets another, the quoted post is nested here as a full tweet object
replyingTo / replyingToID	If this is a reply, the parent's handle and ID
communityNote	Community Note text if one is attached
article	Long-form article body (X Articles), if any
pollData	Poll options and vote counts, if any

Tips

• Quote tweets: check qrt. If non-null, you have a quote tweet — read both text (the wrapper) and qrt.text (the quoted post) to understand the full meaning. Don't summarize a QT without the quoted content.
• Threads: VxTwitter returns a single post, not the full thread. If the user wants a thread, fetch the root post, then walk replies via the x-api skill if installed (conversation_id search) — otherwise ask the user for each URL in the chain.
• Replies: if replyingTo is set, the post may not make sense in isolation. Fetch the parent (replyingToID) for context before drafting your own reply.
• Media: media_extended has richer info (type, dimensions, video bitrates) than mediaURLs. For images you can curl -O directly; for video, pick the highest-bitrate variant from media_extended[].variants.
• Long posts: VxTwitter returns the full text for long-form posts — no note_tweet flag needed (unlike the official API).
• Deleted/protected posts: a 404 or error response means the post is gone, private, or the handle/ID is wrong. Tell the user; don't guess at the contents.
• URL parsing: handle is the segment after the host; tweet ID is the segment after /status/. Strip query strings, fragments, and trailing slashes before constructing the API URL.

Browsing a user's tweets (no auth)

VxTwitter and FxTwitter only do single posts and profile metadata — they do not expose user timelines. Use one of these two paths instead.

Path A — Recent posts (last ~20): Nitter RSS

Best for "what has @handle been posting lately" or "show me their last few tweets".

curl -sS -A "Mozilla/5.0" "https://nitter.net/{handle}/rss"

Returns an RSS feed where each <item> has:

• <title> — full post text (long posts included)
• <link> — https://nitter.net/{handle}/status/{tweet_id}#m (extract the ID)
• <pubDate> — RFC-822 timestamp
• <dc:creator> — author handle (matches retweets vs originals)

Coverage: ~20 most recent posts, typically the last few weeks for an active account. No pagination.

Mirror health (re-tested June 11, 2026): treat Nitter RSS as opportunistic only. nitter.net/{handle}/rss returns 502 Bad Gateway with body stream reset by client (INTERNAL_ERROR) in this environment. xcancel.com/{handle}/rss resolves to rss.xcancel.com and returns an RSS-shaped whitelist stub ("RSS reader not yet whitelisted"), not the user's timeline. nitter.poast.org, nitter.tiekoetter.com, and nitter.privacyredirect.com front with JS/bot-check pages (Anubis / "Making sure you're not a bot!"). Probe nitter.net once; if it returns 5xx or any mirror returns a bot-check/whitelist page, stop trying Nitter for that run and fall back to Path C (syndication) or Path B (search).

To extract tweet IDs from the feed:

curl -sS -A "Mozilla/5.0" "https://nitter.net/{handle}/rss" \
  | grep -oE 'status/[0-9]+' | sort -u

Then enrich any post you want full structured data for (metrics, media, qrt, etc.) by feeding the ID through the VxTwitter pattern at the top of this skill.

Path B — Older or topical tweets: Google site search

Best for "find tweets from @handle about {topic}" — Nitter only sees the recent window, but Google has years of indexed posts.

Use the WebSearch tool with this query shape:

{handle} {topic keywords} site:x.com

Concrete example for the user's "PixelJanitor shadows" case:

PixelJanitor shadows site:x.com

Notes:

• Use site:x.com (not site:x.com/{handle}/status — the narrower path returns no results).
• Results include profile pages, /status/ URLs, and unrelated accounts with similar names — filter by URL prefix https://x.com/{handle}/status/.
• For each /status/{id} URL you find, fetch the full post via VxTwitter to get the actual text (Google snippets are often truncated or stale).

Path C — Top ~100 tweets: Twitter's syndication endpoint

Best for "what are @handle's notable/best tweets" or as the timeline fallback when Nitter is down. This is the official embed-widget backend — no auth:

curl -sS -A "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36" \
  "https://syndication.twitter.com/srv/timeline-profile/screen-name/{handle}" -o timeline.html

Returns a ~500KB HTML page with the timeline embedded as JSON. Extract it:

python3 - <<'EOF'
import re, json
html = open('timeline.html').read()
m = re.search(r'<script id="__NEXT_DATA__" type="application/json">(.*?)</script>', html, re.S)
entries = json.loads(m.group(1))['props']['pageProps']['timeline']['entries']
for e in entries:
    t = e.get('content', {}).get('tweet')
    if t:
        print(t['id_str'], t['created_at'], t['favorite_count'], t['full_text'][:200], sep=' | ')
EOF

Caveats (as of June 2026):

• Returns ~100 tweets sorted roughly by engagement (top tweets), not recency — the pinned tweet comes first. Good for finding someone's best content; it does NOT replace Path A for "what did they post this week."
• Each tweet object includes full_text, favorite_count, in_reply_to_screen_name, and nested retweeted_status/quoted_status — usually no VxTwitter enrichment needed.
• cdn.syndication.twimg.com/timeline/profile?screen_name={handle} (the older JSON variant) returns an empty 200 — use the syndication.twitter.com HTML endpoint above.
• A plain curl/8.x User-Agent may be rejected; send a browser UA.

Recommended workflow for "find good detailed tweets from @handle about {topic}"

1. Pull nitter.net/{handle}/rss and scan the ~20 recent items by topic — fast, no rate limits.
2. Pull the syndication endpoint (Path C) for their ~100 top tweets — catches high-signal older posts that recency and search both miss.
3. If still thin, run WebSearch with {handle} {topic} site:x.com to surface historical posts.
4. Collect candidate tweet IDs from all sources.
5. For each candidate, curl https://api.vxtwitter.com/{handle}/status/{id} to confirm it's real, get the full text, metrics, media, and quote-tweet context.
6. Present the user a short ranked list with URL + 1-line summary + why it matched.

Profile metadata

If you only need bio/follower counts/etc. (not posts), VxTwitter and FxTwitter both have user endpoints:

curl -sS "https://api.vxtwitter.com/{handle}"   # description, followers_count, location, ...
curl -sS "https://api.fxtwitter.com/{handle}"   # similar shape, slightly different fields

When NOT to use this skill

• Authenticated search across all of X, fetching a user's mentions/followers/following, list members, or anything requiring auth → use the x-api skill (requires X_BEARER_TOKEN; if it's not installed, tell the user authenticated operations are unavailable).
• Going beyond ~20 recent posts AND beyond what Google has indexed → only the authenticated API can paginate a user's full timeline.
• Posting, liking, replying, or any write action → not supported by either skill.