Analyze X Post Performance via Browser

Read real engagement data directly from X's interface and run Phoenix scoring analysis — no manual data entry needed.

Input

The user provides one of:

• A post URL — analyze that specific post
• "recent" — analyze the most recent posts from their profile
• Their X username — navigate to their profile to analyze recent posts
• Nothing — ask what they want to analyze

Process

Step 1 — Get Browser Context

mcp__claude-in-chrome__tabs_context_mcp(createIfEmpty: true)

Create a new tab or reuse an existing x.com tab.

Step 2 — Navigate to the Content

For a specific post URL:

mcp__claude-in-chrome__navigate(url: "<post_url>", tabId: <tab>)
mcp__claude-in-chrome__computer(action: "wait", duration: 3, tabId: <tab>)

For recent posts (navigate to profile):

mcp__claude-in-chrome__navigate(url: "https://x.com", tabId: <tab>)
mcp__claude-in-chrome__computer(action: "wait", duration: 2, tabId: <tab>)

Find and click on the user's profile:

mcp__claude-in-chrome__find(query: "profile link or avatar in sidebar", tabId: <tab>)

Verify logged-in state — take a screenshot:

mcp__claude-in-chrome__computer(action: "screenshot", tabId: <tab>)

If not logged in, stop and tell the user to log in first.

Step 3 — Read Post Metrics

#### For a specific post:

Navigate to the post and read its engagement metrics. X shows metrics below each post (replies, reposts, likes, bookmarks, views).

1. Read the page to find metric elements:

mcp__claude-in-chrome__read_page(tabId: <tab>, filter: "all", depth: 10)

1. Or use JavaScript to extract metrics from the post detail page:

mcp__claude-in-chrome__javascript_tool(action: "javascript_exec", text: "
  // Extract metrics from post detail page
  const metrics = {};
  const groups = document.querySelectorAll('[role=\"group\"]');
  const ariaLabels = Array.from(document.querySelectorAll('[aria-label]'))
    .map(el => el.getAttribute('aria-label'))
    .filter(label => label && (
      label.includes('repl') || label.includes('repost') ||
      label.includes('like') || label.includes('bookmark') ||
      label.includes('view') || label.includes('impression')
    ));
  JSON.stringify(ariaLabels);
", tabId: <tab>)

1. If metrics aren't visible or parseable, try clicking the post's analytics/stats icon:

mcp__claude-in-chrome__find(query: "view post analytics or post stats icon", tabId: <tab>)

1. Take a screenshot of the metrics for reference:

mcp__claude-in-chrome__computer(action: "screenshot", tabId: <tab>)

#### For recent posts on profile:

1. Read the profile timeline to get recent posts:

mcp__claude-in-chrome__get_page_text(tabId: <tab>)

1. For each of the last 3-5 posts, extract visible metrics (views, replies, reposts, likes, bookmarks).

1. Use JavaScript to collect metrics from timeline items:

mcp__claude-in-chrome__javascript_tool(action: "javascript_exec", text: "
  const articles = document.querySelectorAll('article');
  const posts = Array.from(articles).slice(0, 5).map((article, i) => {
    const text = article.innerText.substring(0, 100);
    const ariaLabels = Array.from(article.querySelectorAll('[aria-label]'))
      .map(el => el.getAttribute('aria-label'))
      .filter(label => label && /\\d/.test(label));
    return { index: i, preview: text, metrics: ariaLabels };
  });
  JSON.stringify(posts, null, 2);
", tabId: <tab>)

1. Scroll down if needed to load more posts:

mcp__claude-in-chrome__computer(action: "scroll", coordinate: [640, 400], scroll_direction: "down", scroll_amount: 3, tabId: <tab>)

Step 4 — Access Post Analytics Detail (if available)

X Premium users have detailed analytics. Try navigating to the analytics dashboard:

mcp__claude-in-chrome__navigate(url: "https://x.com/analytics", tabId: <tab>)
mcp__claude-in-chrome__computer(action: "wait", duration: 3, tabId: <tab>)
mcp__claude-in-chrome__computer(action: "screenshot", tabId: <tab>)

If analytics is available, extract:

• Impressions over time
• Engagement rate
• Top performing posts
• Follower growth

If not available (not Premium), work with the per-post metrics gathered in Step 3.

Step 5 — Phoenix Score Analysis

With the collected metrics, run the full analysis from the /review skill framework:

Read reference material:

• ../../reference/scoring.md

For each post analyzed, produce:

Performance Table:

Key Ratios:

• Impressions / followers: >2× means algorithm is amplifying
• Reply-to-like ratio: >0.15 = strong conversation driver
• Bookmark-to-like ratio: >0.10 = high-quality content signal

Conversation Velocity Assessment:

• Did the author reply to comments? (check reply threads on the post)
• How many author reply threads exist?
• Were replies substantive or just "thanks"?

Diagnosis:

• Why did this post perform as it did?
• What specific elements drove engagement (or didn't)?
• What would you change for next time?

Step 6 — Comparative Analysis (if multiple posts)

If analyzing multiple recent posts, produce:

1. Ranked performance table — posts sorted by engagement rate
2. Pattern identification — what formats/topics/times performed best
3. Trend line — are things improving, declining, or flat?
4. Top recommendation — one specific action to improve the next post

Step 7 — Output

Present:

1. Raw metrics collected from X (with screenshot reference)
2. Phoenix score analysis per post
3. Key ratios with benchmarks
4. Diagnosis — why each post performed as it did
5. Recommendations — specific actions for the next post
6. Suggested next content — based on what's working, suggest the next post topic/format

Offer to run /compose on the suggested next content.

Tips

• X shows different metric granularity based on Premium status
• The analytics page (x.com/analytics) requires Premium
• Per-post metrics (reply/repost/like/bookmark counts) are visible to everyone
• View counts are visible on most posts
• For the most accurate data, analyze posts that are at least 24 hours old (let the distribution curve complete)

Troubleshooting

• Metrics not loading: Wait longer (3-5 seconds), X lazy-loads engagement counts
• Can't find metrics: Take a screenshot and visually identify where numbers appear
• Analytics page redirects: User may not have Premium — fall back to per-post metrics
• JavaScript extraction fails: Fall back to read_page with filter: "all" and search for numeric elements
• Rate limiting: If X shows "something went wrong", wait 30 seconds and retry once