tweet-content

name	tweet-content
description	Pick a random item from the trading content collection and compose a tweet about it on X.
disable-model-invocation	true

Tweet trading content

Pick a random item from the algorithmic trading content collection (papers, blog posts, AI/ML resources) and compose a tweet about it on X (Twitter).

Steps

1. Check browser is activated:

   • Call tabs_context_mcp to verify browser connection
   • If it fails, ask the user to enable browser:
     • VS Code: Ensure Chrome extension is connected
     • CLI: Run claude --chrome or type /chrome

2. Extract all tweetable items from RST files: Read the following RST files and extract every entry (title, description, URL, author if present, and any "Mentioned by" discussion links):

   • source/learn/papers.rst
   • source/learn/ai-and-machine-learning.rst
   • source/learn/blog-posts.rst

   For each entry, extract:

   • title: The RST section title
   • url: The link URL from the Read the paper/post/... line
   • description: The description paragraph(s)
   • author: Author name if present (lines starting with "By ")
   • mention_url: Any Twitter/X discussion URL from "Mentioned by" lines
   • mention_author: The person who mentioned it
   • mention_quote: What they said about it
   • source_file: Which RST file it came from

   Write the full list as a Markdown file to the scratchpad directory at: /private/tmp/claude/-Users-moo-code-docs/882c5ece-843e-4dd4-ab2c-94ca5a77e61c/scratchpad/tweetable-items.md

   Format each item as:

   ## Title Here
   - **URL**: https://...
   - **Source**: papers.rst | ai-and-machine-learning.rst | blog-posts.rst
   - **Author**: Name (if available)
   - **Discussion**: [Mentioned by X](https://x.com/...) (if available)
   - **Description**: Brief description...
   

3. Pick a random item: Select one item at random from the list. Tell the user which item was selected.

4. Capture the article screenshot: Use browser MCP to navigate to the article URL and save a screenshot. Steps:

   • Use navigate to open the article URL
   • Wait for page to load; if CAPTCHA appears, ask the user to complete it manually
   • Use computer with action screenshot to capture the page
   • Save the screenshot to disk using javascript_tool:

     // Get the screenshot as a data URL and trigger a download
     // Alternatively, use the returned screenshot data from the computer tool
     

   • Save the image to the scratchpad: /private/tmp/claude/-Users-moo-code-docs/882c5ece-843e-4dd4-ab2c-94ca5a77e61c/scratchpad/tweet-image.png
   • If saving fails, use Bash: screencapture or fall back to a manual approach

   If the page cannot be accessed (paywall, CAPTCHA), tell the user and proceed without an image.

5. Look up author Twitter handle: If the item has a known author, look up their Twitter/X handle:

   • Read source/learn/influencers.rst and match the author name to a listed handle
   • If the author is not in influencers.rst, use browser to check the article page for a Twitter handle clue (e.g., an X.com link in the post, or the author's profile). If found, add the author to the appropriate section of source/learn/influencers.rst
   • Store the handle (e.g., @pedma7) for use in the tweet

6. Create tweet text: Write a concise two-line summary of the item suitable for a tweet. Guidelines:

   • First line: What the paper/post is about and why it matters (the hook)
   • Second line: A key finding or insight
   • Attribute the author with "By @handle" (using the Twitter handle from step 5)
   • Add the URL at the end
   • If the item has a "Mentioned by" discussion, add a line like: Discussed by @handle here: <discussion_url>
   • Keep the total under 280 characters (excluding the URL which X auto-shortens to 23 chars)
   • Use a professional but engaging tone appropriate for fintwit/quant twitter

   Save the tweet text to the scratchpad: /private/tmp/claude/-Users-moo-code-docs/882c5ece-843e-4dd4-ab2c-94ca5a77e61c/scratchpad/tweet-text.txt

7. Open X compose and prepare the draft in browser: Use browser MCP to navigate to X and fill in the tweet compose form with text and image.

   a. Navigate to X compose:

   • Use navigate to open https://x.com/compose/post
   • Wait for the compose dialog to load
   • If a login page or CAPTCHA appears, ask the user to handle it, then continue

   b. Type the tweet text:

   • Use find to locate the tweet text input area (the contenteditable div in the compose dialog)
   • Use computer with action click to focus the text input
   • Use computer with action type to type the tweet text character by character
   • Alternatively, use javascript_tool to set the text directly:

     (() => {
       const editor = document.querySelector('[data-testid="tweetTextarea_0"] [contenteditable="true"]');
       if (!editor) return 'ERROR: Could not find tweet editor';
       editor.focus();
       // Use execCommand to trigger React's change detection
       document.execCommand('insertText', false, TWEET_TEXT_HERE);
       return 'OK';
     })();
     

   • Verify the text appears correctly using computer with action screenshot

   c. Attach the screenshot image (if one was captured in step 4):

   • Use find to locate the media upload button (the image/photo icon in the compose toolbar)
   • Use computer with action click on the media upload button
   • This opens a file picker dialog — use form_input or computer to interact with it

8. Stop and wait for user approval: Tell the user:

   • "The tweet draft is ready in your browser with text and image attached."
   • "Please review the compose window and click Post when ready."
   • If the image could not be attached automatically: "The screenshot path has been copied to your clipboard — use the media button in the compose window to attach it manually." Do NOT click the Post button. The user must post manually.