Description
Project overview I run a tutoring business. We have a working automation that monitors a Slack channel for tutor-match requests, extracts calendar availability data from TutorBird (behind a login), cross-references a Google Sheet roster with a lot of data for each tutor, and posts a ranked list of matching tutors back to Slack as a threaded reply. The system currently runs on Browser-Use's cloud platform and is unreliable — it's expensive and often breaks, and I want to move to a more predictable stack. I have complete, detailed spec files for the existing workflow that will inform the rebuild. What you'll be building A production-grade workflow orchestrated in n8n (I'll pay for n8n Cloud) that: Triggers in real time via Slack's Events API (no cron polling) — the workflow starts within seconds of a new message in the target channel Filters eligible messages as the first step (regex + reaction-based dedupe); ineligible messages exit immediately For each eligible request: Reads a tutor roster from a Google Sheet Triggers a Skyvern workflow to log into TutorBird and extract tutor availability windows for the relevant days Calls the Anthropic API (Claude Sonnet 5) with a matching prompt to rank tutors against the client's requirements Posts the ranked result as a threaded Slack reply Manages reaction state on the original message (🔄 while processing → ✅ on success, ❌ on failure) Handles error cases and dedupes cleanly Includes a lightweight hourly safety-net workflow that scans