Produces a caching strategy for a system — goals, what to cache ranked by impact, cache layers, a per-resource strategy (TTL, invalidation, key pattern), failure modes including stampede and staleness, monitoring, and a cold-start plan. Use when the user says "caching strategy", "what should we cache", "reads are slow", "add Redis/CDN", or describes read-latency pain. Use this to fix read latency with caching; use scalability-assessment to find which component is the bottleneck first, and database-selection-guide when the real fix is a different datastore.
