xclaw

Extract posts from X (Twitter) timeline or profile pages with engagement metrics, media URLs, and download local copies. Use when asked to scrape, list, or a...

Security Scan

The skill's instructions, file writes, and scraping behavior match its stated purpose (scraping X/Twitter timelines and saving media) and do not request unrelated credentials or installs.

Purpose & Capability

Name/description align with the runtime instructions: the SKILL.md describes DOM scraping of X pages, collecting engagement metrics and media, and saving results to a defined folder structure. No unrelated binaries, env vars, or installs are requested.

Instruction Scope

Instructions tell the agent to navigate a browser, run page-evaluation JS to extract posts, create folders (../../intel/x/media/...), and download media. This is expected for a scraper, but the skill implicitly reads whatever the browser session can access (including a logged-in home timeline and any session cookies) and writes files to relative paths two levels up—so there is a privacy risk if the agent is logged into a personal account or the target page contains private data.

Install Mechanism

No install spec and no code files — instruction-only. This minimizes supply-chain risk because nothing is downloaded or written at install time.

Credentials

The skill requires no environment variables or credentials, which is coherent. However, it assumes access to a browser context that may be authenticated; that implicit access to session cookies is not declared and could expose private timeline data when run.

Persistence & Privilege

always is false and there is no install-time persistence. The skill does instruct file creation within the agent workspace (or parent directories) but does not request elevated or persistent platform privileges.

Scan Findings in Context

[no_findings] expected: The regex scanner found no code patterns; this is expected because the skill is instruction-only and contains only SKILL.md.

Assessment

This skill appears to do what it says: it will navigate X pages, scrape DOM content, and save an intel markdown file plus media into ../../intel/x/media/{timestamp}. Before installing or running it, consider: (1) If your agent's browser is logged into your X account, the skill may scrape private/home-timeline content — log out or use an unauthenticated session if you don't want that. (2) The skill will write files two levels up from the execution directory—verify that path is acceptable and won't overwrite important files. (3) Scraping may violate X/Twitter terms of service or copyright; review legal/ethical implications. (4) If you need the skill to access protected data (e.g., private messages or an authenticated feed), expect to provide authentication separately; the skill does not declare credentials. If these points are acceptable, the skill is internally coherent.

Extract posts from X (Twitter) pages with full metadata including text, author, timestamp, engagement metrics, media URLs, and download local copies.

When to Use

• List recent posts from a user's profile
• Extract posts from home timeline
• Capture post URLs without navigating into individual posts
• Gather engagement metrics (likes, reposts, replies, views)
• Download images and video thumbnails
• Analyze content patterns across multiple posts

---

File Structure

intel/x/
├── {date}-{time}-X.md      # Intel file
└── media/
    └── {YYYYMMDD}-{HHMMSS}/        # Media folder (same timestamp)
        ├── image_{postId}_{index}.jpg
        └── video_{postId}_poster.jpg

---

File Naming

Item	Format	Example
Intel file	{date}-{time}-X.md	2026-03-16-143052-X.md
Media folder	{YYYYMMDD}-{HHMMSS}	20260316-143052
Image	image_{postId}_{index}.jpg	image_2032741269689213289_0.jpg
Video thumbnail	video_{postId}_poster.jpg	video_2032741269689213289_poster.jpg

---

OUTPUT FORMAT — EXACT FORMAT REQUIRED

WRITE THE INTEL FILE IN THIS EXACT FORMAT. FOLLOW EVERY COMPONENT PRECISELY.

# X Intelligence Report

**Scraped:** {date} {time} (Asia/Shanghai)
**Source:** X Home Timeline
**Posts Collected:** {count}

---

## Post {n}: {Topic Title}

**Author:** {Name} (@{handle})
**Posted:** {relative time}
**URL:** {post_url}

**Content:**
{post_text_content}

**Metrics:** ❤️ {likes} 🔁 {reposts} 💬 {replies} 👁 {views}

**Media:**
🖼️ media/{folder}/image_{postId}_{index}.jpg

**Signal:** {🔴 HIGH / 🟡 MEDIUM / 🟡 LOW / ⚪ SKIP} — {Brief signal assessment}

---

## Post {n+1}: ...

(MORE POSTS...)

---

## Summary

**High Signal (Tier 1):**
1. {Topic} — {One-line summary}
2. {Topic} — {One-line summary}

**Medium Signal (Tier 2):**
- {Topic} — {One-line summary}
- {Topic} — {One-line summary}

**Skip:**
- {Topic} — {Reason}

Media in Output Format (Conditional)

IF THE POST HAS MEDIA, INCLUDE THE MEDIA LINE AFTER METRICS:

**Media:**
🖼️ media/{folder}/image_{postId}_{index}.jpg

OR FOR VIDEO THUMBNAILS:

**Media:**
🎬 media/{folder}/video_{postId}_poster.jpg

IF THE POST HAS NO MEDIA, OMIT THE MEDIA SECTION ENTIRELY.

---

WORKFLOW

Step 1: Create Media Folder FIRST

BEFORE EXTRACTING ANY POSTS, create the media folder:

# Create folder with current timestamp
# Format: YYYYMMDD-HHMMSS
mkdir -p ../../intel/x/media/{YYYYMMDD}-{HHMMSS}

# Example
mkdir -p ../../intel/x/media/20260316-143052

Step 2: Navigate to Target Page

// For home timeline
browser.navigate({ url: "https://x.com/home" })

// For a specific user profile
browser.navigate({ url: "https://x.com/username" })

Step 3: Scroll to Load More Posts

browser.act({ 
  fn: "() => { for(let i=0; i<3; i++) { window.scrollBy(0, window.innerHeight); } return 'scrolled'; }",
  kind: "evaluate"
})

Step 4: Extract Posts with Media URLs

browser.act({
  fn: "() => {
    const articles = document.querySelectorAll('article[data-testid=\"tweet\"]');
    const posts = [];
    articles.forEach(article => {
      const url = article.querySelector('a[href*=\"/status/\"]')?.href;
      if (!url || url.includes('/analytics') || url.includes('/photo')) return;
      
      const text = article.querySelector('[data-testid=\"tweetText\"]')?.innerText || '';
      const timeEl = article.querySelector('time');
      const timestamp = timeEl?.innerText || '';
      const datetime = timeEl?.dateTime || '';
      
      // Get author info
      const nameEl = article.querySelector('[data-testid=\"User-Name\"]');
      const name = nameEl?.querySelector('span')?.innerText || '';
      const handle = nameEl?.querySelector('a[href*=\"/\"]')?.innerText || '';
      
      // Get engagement metrics
      const metricButtons = article.querySelectorAll('button[role=\"button\"]');
      let replies = '0', reposts = '0', likes = '0', views = '0';
      
      metricButtons.forEach(btn => {
        const txt = btn.innerText || '';
        const label = btn.getAttribute('aria-label') || '';
        if (label && label.includes('Reply')) {
          const match = txt.match(/(\d+)/);
          if (match) replies = match[1];
        }
        if (label && label.includes('Repost')) {
          const match = txt.match(/(\d+)/);
          if (match) reposts = match[1];
        }
        if (label && label.includes('Like')) {
          const match = txt.match(/(\d+)/);
          if (match) likes = match[1];
        }
      });
      
      // Views are in an <a> link (not button), e.g. <a href="/.../analytics">71K</a>
      const viewLink = article.querySelector('a[href*=\"/analytics\"]');
      if (viewLink) {
        const v = viewLink.innerText.trim();
        views = v.replace(/[^0-9KMB.]/g, '') || '0';
      }
      
      // Get media URLs - images and video thumbnails
      const mediaUrls = [];
      
      // Image posts: pbs.twimg.com/media/
      const images = article.querySelectorAll('img[src*=\"pbs.twimg.com/media/\"]');
      images.forEach(img => {
        if (img.src && !mediaUrls.includes(img.src)) {
          mediaUrls.push(img.src);
        }
      });
      
      // Also check for image links in <a> tags
      const imageLinks = article.querySelectorAll('a[href*=\"pbs.twimg.com/media/\"]');
      imageLinks.forEach(link => {
        if (link.href && !mediaUrls.includes(link.href)) {
          mediaUrls.push(link.href);
        }
      });
      
      // Video thumbnails: pbs.twimg.com/amplify_video_thumb/
      const videoThumbs = article.querySelectorAll('img[src*=\"pbs.twimg.com/amplify_video_thumb/\"]');
      videoThumbs.forEach(img => {
        if (img.src && !mediaUrls.includes(img.src)) {
          mediaUrls.push(img.src);
        }
      });
      
      const postId = url.match(/\/status\/(\d+)/)?.[1] || '';
      
      posts.push({ 
        name,
        handle,
        text: text.substring(0, 500),
        timestamp, 
        datetime,
        url,
        postId,
        mediaUrls,
        metrics: { replies, reposts, likes, views }
      });
    });
    return posts.slice(0, 10);
  }",
  kind: "evaluate"
})

Step 5: Download Media Files

FOR EACH POST WITH MEDIA, DOWNLOAD THE FILES:

# Download image - replace URL with actual media URL from extraction
curl -L -o "../../intel/x/media/{folder}/image_{postId}_{index}.jpg" "https://pbs.twimg.com/media/xxx?format=jpg&name=large"

# Download video thumbnail
curl -L -o "../../intel/x/media/{folder}/video_{postId}_poster.jpg" "https://pbs.twimg.com/amplify_video_thumb/xxx"

IMPORTANT:

• USE -L FLAG TO FOLLOW REDIRECTS
• ADD &name=large FOR FULL-RESOLUTION IMAGES
• USE THE ACTUAL URLS EXTRACTED FROM STEP 4

Step 6: Write Intel File

WRITE THE MARKDOWN FILE TO ../../intel/x/{date}-{time}-X.md

FOLLOW THE EXACT OUTPUT FORMAT SHOWN ABOVE. INCLUDE:

• Header with Scraped, Source, Posts Collected
• Each post with Author, Posted, URL, Content, Signal
• Media references if present
• Summary section at the end

Step 7: Close Chrome Tab

browser.close()

---

Key Selectors

Element	Selector
Article container	article[data-testid="tweet"]
Post text	[data-testid=\"tweetText\"]
Timestamp	time
User name	[data-testid=\"User-Name\"]
Post URL	a[href*=\"/status/\"]
Analytics link (views)	a[href*=\"/analytics\"]
Metric buttons	button[role=\"button\"]
Images	img[src*=\"pbs.twimg.com/media/\"]
Video thumbnails	img[src*=\"pbs.twimg.com/amplify_video_thumb/\"]

---

Tips

1. Don't click into posts - Extract URLs directly from the timeline
2. Filter URLs - Exclude /analytics and /photo URLs
3. Scroll before extracting - X loads posts dynamically
4. Create folder FIRST - Media folder must exist before downloads
5. Use -L flag - curl must follow redirects for pbs.twimg.com URLs
6. Add &name=large - Get full-resolution images, not thumbnails
7. Always close the tab - Task ends when tab closes