X Reader

Use the bundled Node script to read X links with minimal token overhead.

Workflow

1. Run scripts/xreader.mjs with the URL.
2. Let the script decide the cheapest path:
   • normal tweet → xreach tweet
   • explicit thread request → xreach thread
   • article / t.co → resolve first, use Playwright only when needed
3. Return the script's structured JSON; summarize from that instead of pasting raw page content.

Positioning

• No official X API required.
• No developer application or paid X API access required.
• Works with login cookies plus low-token routing, which keeps normal tweet reads cheap and uses Playwright only when article rendering is necessary.
• Compared with xreach, this skill is a higher-level reader for agent workflows: it adds article handling, authored-thread filtering, and unified structured output.
• Compared with xcurl, this skill focuses on content extraction from links instead of low-level request control.

Commands

Summary mode (default):

node skills/x-reader/scripts/xreader.mjs "https://x.com/..."

Full mode:

node skills/x-reader/scripts/xreader.mjs --mode full "https://x.com/..."

Thread mode (explicit only, to save tokens):

node skills/x-reader/scripts/xreader.mjs --thread "https://x.com/.../status/..."

Debug mode (headed browser for article extraction):

node skills/x-reader/scripts/xreader.mjs --debug "https://x.com/i/article/..."

Dependencies

Required:

• xreach
• Node.js
• valid X auth cookies

Install article-mode dependency inside the skill directory:

cd skills/x-reader
npm install

Auth

Primary auth path:

~/.config/xreader/session.json

Legacy fallback path:

~/.config/xfetch/session.json

If legacy auth exists, the script migrates it to the new path automatically.

Expected JSON format:

{
  "authToken": "...",
  "ct0": "..."
}

Output contract

Expect structured JSON with fields such as:

• ok
• type (tweet, thread, or article)
• url
• canonicalUrl
• source (xreach or playwright)
• author
• publishedAt
• title
• text / contentText
• contentMarkdown
• summaryText
• fallback / warnings / error

Prefer quoting or summarizing summaryText for low-token responses. Use contentText or contentMarkdown only when the user clearly wants more detail.

Quick verification

Run these after install/auth setup:

node skills/x-reader/scripts/xreader.mjs "https://x.com/yangguangai/status/2033736815405121642?s=46"
node skills/x-reader/scripts/xreader.mjs "https://x.com/yangguangai/status/2033522959407878621?s=46"
node skills/x-reader/scripts/xreader.mjs --thread "https://x.com/google/status/2031558824042058064"

Notes

• Default tweet reads are single-post only for token efficiency.
• Use --thread only when the user clearly wants the authored thread.
• Article extraction is MVP quality: good enough for reading and summarization, but long X articles with many examples/code blocks may still include some template noise.
• If article extraction fails, the script falls back to the original tweet payload and marks the failure in warnings / articleError.
• This skill relies on X login cookies (auth_token + ct0). Cookie-based automation may carry account risk, including additional verification or account restrictions. The risk is usually low for light personal use, but recommend using a secondary account instead of a primary high-value account.

Release checklist

• Confirm SKILL.md includes usage, dependencies, auth path, and limitations.
• Confirm node skills/x-reader/scripts/xreader.mjs --help works.
• Smoke test one tweet, one article, and one thread URL.
• Confirm auth file is not packaged into the skill.
• Package the skill and verify validation passes before publishing.