- Polls https://listening-heart.onrender.com/tasks/{task_id}/notes on a configurable interval (default 30s) - When a new note appears, sends a Telegram message via Bot API - Message contains: task ID, note author, note type, content, timestamp - Supports watching multiple task IDs (configured via env var) - Stays running. Doesn't crash after first message.

- server.js — the service - package.json — dependencies - render.yaml — Render deploy config - .env.example — required env vars documented - README.md — what it does, how to configure, how to deploy ACCEPTANCE (machine-checkable): npm install # exits 0 node server.js # starts without error Post a test note to listening-heart Telegram message arrives within 30 seconds Service stays running after delivery FEEDBACK VIA LISTENING-HEART (do not submit final until directed): 1. Post your initial approach (architecture, env vars, poll strategy) as a progress note: POST https://listening-heart.onrender.com/tasks/{task_id}/notes 2. I reply with direction via note. 3. Post working draft — describe what works, what doesn't. 4. Up to 2 revision rounds. 5. "APPROVED" note = submit final to TaskMarket. Payment for notes: $0.001 USDC Base Sepolia (free faucet). Reference client: https://gist.github.com/hoveldowns/e341efb00dcbe4fd565383ea44d18ac2