feat: enhanced scrapeLikedTweets scraper with rich data extraction

[nj-io]

Summary

Replaces the broken xeepy-based x_get_likes handler with a proper scrapeLikedTweets() scraper, following the same pattern as scrapeBookmarks().

What changed

• New scrapeLikedTweets() in src/scrapers/twitter/index.js — scroll-based scraper with rich data extraction
• Wired through src/scrapers/index.js → src/mcp/local-tools.js (same pattern as bookmarks)
• Removed x_get_likes from xeepyTools array and deleted the broken executeXeepyTool handler that called the non-existent localTools.getPage()

Rich data per tweet

Field	Source
text	Full text with "Show more" expansion
author, handle	User-Name + first a[href]
timestamp, link	time[datetime], first /status/ link
images	a[href*="/photo/"] attributed to correct author by handle matching
quotedTweet	Detected via multiple UserAvatar-Container-* elements (X doesn't always use quoteTweet testid)
article	article-cover-image + nextElementSibling for title/description
card	card.wrapper for link previews
replies, retweets, likes, views	Parsed from role="group" aria-label

Bug fixes

• "Show more" clicks one at a time — X re-renders the DOM after each click, detaching all other button references. Previous approach of collecting all buttons then iterating caused Node is detached from document errors.
• Article URL construction — only builds article.url for direct articles (not quoted tweets where the article belongs to a different author). Always includes article.tweetUrl for reliable resolution.

Relation to #7

This overlaps with the scrapeLikedTweets portion of #7. The scraper here is more comprehensive (quote tweets, articles, cards, engagement stats, image attribution) vs #7's basic fields. Both follow the same architectural pattern requested by the maintainer: scraper in twitter/index.js, routed through local-tools.js, removed from xeepyTools.

Test plan

• x_get_likes returns rich data with quote tweets, articles, cards, engagement stats
• "Show more" expansion works for truncated tweets
• Scrolling collects beyond the initial viewport
• Deduplication prevents duplicate tweets
• Existing tools unaffected (no changes to other handlers)

🤖 Generated with Claude Code

[vercel]

@nj-io is attempting to deploy a commit to the kaivocmenirehtacgmailcom's projects Team on Vercel.

A member of the Team first needs to authorize it.

[nj-io]

Superseded by updated PR with timestamp filtering, auth checks, and proper scraper pattern.