Weekly Product Hunt discovery digest for product teams

Every Monday at 9am, run a deterministic Product Hunt discovery harvest for my product and growth team. This is the same-pages, same-filter, same-format digest each week — no reasoning inside the loop.

Kick off an Airtop cloud browser session and use Load URL to open Product Hunt's weekly leaderboard at https://www.producthunt.com/leaderboard/weekly. Then use Paginated Extraction to pull every launch on the leaderboard, capturing for each one: product name, tagline, the category tags Product Hunt has assigned, the current upvote count, the maker names, and the launch URL. Terminate the Airtop session as soon as extraction finishes so the run doesn't keep billing credits.

Once the raw list is in memory, filter it down to launches that touch my tracked topics. A launch qualifies if any of its category tags, its name, or its tagline contains any of the tracked keywords (case-insensitive substring match). Default the tracked topic list to AI agents, dev tools, and sales tech, and expose it as a workflow input the user can edit.

For every qualifying launch, call Google Sheets Append Values against my Product Hunt tracker spreadsheet, adding a new row with: launch date (the Monday the digest ran, ISO format), name, tagline, matched topic, upvote count, maker names, category tags (joined with commas), and launch URL. Use USER_ENTERED so the URL renders as a clickable link and the date parses as a date.

Finally, call Slack Bot Send a Message to post to #product-radar with the top qualifying launches formatted as a ranked list. Sort by upvote count descending, cap at ten items, and render each entry in Slack mrkdwn as `1. *<launch URL|Product name>* — tagline (N upvotes, matched: topic)`. Lead the message with a one-line header naming the week (for example 'Product Hunt radar — week of Jul 14'). If nothing qualified this week, post a short 'no matches this week' note instead of skipping the message so the team knows the workflow still ran.

This is a deterministic weekly harvest — every step is a known operation over structured input: Airtop Load URL and Paginated Extraction to scrape, an in-workflow keyword filter, Google Sheets Append Values to log, Slack Bot Send a Message to broadcast. Cron trigger; no polling, no LLM in the loop.