What problem does it solve? Inbound provider webhooks (marketplace orders, payment status, WhatsApp messages) must be acknowledged fast, retried safely, and never silently lost — but hand-rolling receivers per channel leads to inconsistent retry logic, lost events, and providers disabling topics after 5xx responses. This Skill documents the shared enqueue-first notification pipeline every channel in this monorepo uses, so new channels inherit identical resilience behavior. ## Core Features & Use Cases - Shared pipeline architecture: Explains defineNotificationPipeline, the disposition matrix (resolve/drop/park/fail/defer), the failures-only Firestore store, and the durable-cursor reprocess sweep. - Two retry lanes: Distinguishes the hot failed lane (1h window, 5 attempts) from the slow deferred lane (24h window, 7 attempts) for preconditions only a human can clear. - Step-by-step channel onboarding: Walks through schema, collection handle, composite index, adapter, receiver, Cloud Tasks scheduler, and Cloud Functions triggers for adding a new channel. - Use Case: When adding a webhook receiver for a new marketplace, follow the guide to wire signature verification, enqueue-first ack-200 handling, idempotent processing, and the reprocess sweep without persisting every event. ## Quick Start Ask the assistant to add a webhook notification receiver for a new channel following the webhook-notifications pipeline, starting with the schema and Firestore composite index.