Turn your X bookmarks into a Notion knowledge base

Build me an agent workflow that turns my X (Twitter) bookmarks into an organized Notion knowledge base. The judgement on summarization, tag selection, category, and thread detection is what makes this an agent workflow rather than a code one.

Trigger: a cron schedule that runs every morning at 7am in my local timezone.

On each run, the agent should do the following:

1. Call X (Twitter) Get Bookmarks to fetch the authenticated user's currently saved bookmarks. Paginate through all results.

2. Call Notion Query a Data Source against my Bookmarks database to read the tweet URLs (or tweet IDs) that are already filed. Use this list to compute the set of new bookmarks that have not been processed yet. Do not reprocess bookmarks that already exist in Notion.

3. For each new bookmark, the agent should:

a. If the bookmark response is missing the full tweet text, author handle, or other context the agent needs, call X Get Tweet on the tweet ID to fetch the full body and author.

b. Write a one or two sentence summary of what the tweet actually says or argues.

c. Choose one or more topical tags from a short configurable list (default starter list: AI, startups, design, frameworks, research, marketing, engineering, personal). The list of tags should be easy for me to edit in the workflow instructions.

d. Pick exactly one Category from a fixed configurable list (default: How-to, Opinion, Resource, Announcement, Thread, Other).

e. Decide whether the tweet is part of a thread worth unrolling, and set a Thread boolean accordingly. Signals to use: the tweet replies to itself, is numbered (1/, 2/), or ends with 🧵.

4. Call Notion Create a Page in my Bookmarks database with these properties: Title (a short agent-written title, not the full tweet), Summary, Tags (multi-select), Category (select), Author (the author's @handle), Tweet URL, Saved Date (the bookmark or tweet timestamp), and Thread (checkbox).

5. In the page body, include two sections: the original tweet text exactly as posted, and a short Notes block where the agent can add why this is worth coming back to (one to three bullets is fine). If Thread is true, add a line that says 'Worth unrolling — open original for full thread.'

Error handling: if Get Tweet returns 404 (deleted) or 403 (restricted), still create the Notion page with whatever metadata is available from the bookmarks response, mark Category as Other, and note 'Original tweet no longer accessible' in the body.

Database assumption: I will share the existing Notion Bookmarks database ID and column schema when configuring the workflow. The agent should be tolerant of small column-name variations and ask me to confirm the mapping on first run if anything looks ambiguous.