Daily Twitter brand mention triage to Notion and Slack

Build an agent workflow that triages Twitter/X brand mentions every weekday morning and ends with a single Slack digest. Treat the brand name and the product keyword list as configurable inputs at the top of the agent instructions.

Trigger: cron, every weekday at 8:30am in the workspace's local time zone.

Step 1. Scrape the last 24 hours of tweets. Call Apify's "Run Actor Synchronously and Get Dataset Items" against a Twitter/X scraper Actor (default to apify~twitter-scraper, but let the user pick the Actor by ID or username~name in the agent inputs). Build the search query from the brand name plus each product keyword joined with OR. Pass a since/until window for the last 24 hours and a reasonable maxItems cap (default 200). Return the dataset items directly. If the Actor returns zero items, log it and skip steps 3 and 4 for the day.

Step 2. For each tweet, classify it into exactly one of: complaint, feature_request, praise, or noise. Drop everything classified as noise. For the rest, assign an urgency score from 1 (low) to 5 (critical). Bump urgency for: account-blocking or outage language, viral threads (high reply/retweet counts), replies coming from large accounts (follower count above a threshold the agent picks sensibly, e.g. 10k), and threads where the original tweet is from a paying-customer signal in the bio. Produce a one-line summary for each kept tweet (max 140 chars).

Step 3. For every kept tweet, call Notion's "Create a Page" against a brand-mentions database. The user supplies the database ID at setup. Set properties for: Author (the handle), Sentiment (complaint / feature_request / praise), Category (free-text, e.g. billing, onboarding, mobile), Urgency (1-5), Link (the tweet URL), and Summary (the one-line summary). Use the summary as the page title.

Step 4. Post a single Slack digest via Slack Bot's "Send a Message" to the channel the user picks at setup. Format: a one-line header with today's date and total relevant mentions, then the top 3 items ranked by urgency (each as a bullet with author, sentiment, summary, and the Notion page link), then a footer line counting the rest by category (e.g. "+12 more: 7 complaints, 3 requests, 2 praise"). Use Slack mrkdwn formatting. Do not post if there were zero relevant mentions.

Inputs the agent should expose at setup: brand name (string), keyword list (array of strings), Apify Actor ID (default apify~twitter-scraper), Notion database ID, Slack channel ID, optional follower-count threshold for the large-account urgency bump.

Error handling: if Apify returns a non-success run, post a single Slack alert to the same channel naming the failure and stop. If Notion page creation fails for one item, log and continue; do not block the rest of the run. Always finish by posting the Slack digest, even if some Notion writes failed (note any failures in the footer).