What problem does it solve? Backend queues and webhook pipelines silently accumulate stuck processing rows, failed deliveries, duplicate provider events, and poison jobs that retry forever. This Skill audits every queue and webhook surface through eight deterministic gates and produces a review-only repair report instead of blindly retrying production jobs. ## Core Features & Use Cases - Eight-gate queue audit: Inventories queue tables, edge functions, and pg_cron jobs, then classifies stuck processing rows, old pending jobs, failed webhook deliveries, duplicate provider event IDs, poison jobs, provider outage patterns, and rows missing tenant_id. - Safety-first repair proposals: Generates review-only SQL migrations ordered as dedupe indexes, quarantine tables, poison-job moves, stuck-row reaping, backoff fixes, and cron repairs, with hard refusals on bulk retries, deletes, and tenant-less row replays. - Use Case: A Telnyx webhook consumer shows duplicate event IDs and an AI invoice queue has rows stuck in processing for hours. Run this Skill to get a full report with redacted samples, root-cause diagnosis, and an idempotent repair migration plan awaiting approval. ## Quick Start Ask the assistant to audit all queue and webhook tables for stuck jobs, failed deliveries, and duplicate event IDs and produce a repair report.
