What problem does it solve? Convex applications can become slow or expensive due to read amplification, OCC write conflicts, excessive reactive subscriptions, or functions hitting execution and transaction limits. This Skill provides a structured audit workflow that starts from real signals (like npx convex insights --details), routes each symptom to the right problem class, and applies targeted fixes without over-engineering. ## Core Features & Use Cases - Signal-driven diagnosis: Gathers evidence from Convex deployment insights, CLI output, or code audits, then routes symptoms to one of four problem classes: hot-path reads, OCC conflicts, subscription cost, or function budget. - Reference-backed fixes: Four detailed reference files cover index usage, denormalization, digest tables, hot document splitting, no-op write elimination, point-in-time reads, and batched mutations, each with concrete code examples and a recommended fix order. - Sibling consistency checks: After fixing one function, the workflow audits sibling readers and writers on the same tables so the same pattern is fixed everywhere. - Use Case: Your dashboard shows high bytes read on a project list query. The Skill traces the query's full read set, replaces a scan-plus-filter with an indexed digest table read, adds fallback logic for unbackfilled fields, and verifies sibling list queries follow the same pattern. ## Quick Start Ask the agent to audit your Convex app for performance issues, starting with the strongest available signal such as insights output or a slow page, and to propose the smallest high-impact fix first.
