What problem does it solve? Choosing the wrong cache backend or missing invalidation logic causes stale data, unbounded memory growth, and rate limits that break under horizontal scaling. This Skill provides a decision framework for matching cache backends to durability needs and enforcing consistent TTL, key naming, and invalidation contracts. ## Core Features & Use Cases - Backend Selection Guidance: Maps durability needs to the right store — in-memory Map for ephemeral run-local deduplication, node-cache for process-local TTL, PostgreSQL for durable shared data, and Redis as the planned distributed target. - TTL and Key Naming Standards: Defines TTL guidelines per data type (7 days for expensive LLM output, 5 minutes for pair scores) and collision-resistant key naming patterns. - Invalidation and Scaling Guardrails: Documents invalidation contracts wired to write paths and flags in-memory caches with TODO(redis) markers for horizontal-scaling readiness. - Use Case: When a user reports stale match explanations after a profile update, use this Skill to trace the invalidation contract and fix substring-based cache deletion that misses roster-hash keys. ## Quick Start Ask the AI to review the caching approach for a new feature, for example: add caching for generated event themes with the right backend, TTL, and invalidation on roster changes.
