/cache

Analyzes caching strategies and implementations, identifies inefficiencies in hit rates, TTL, invalidation, and generates optimization report with code examples.

Diagnoses DeepSeek prefix cache churn by fetching a proxy alignment report and env var state, then explains root causes and next steps.

Designs multi-layer caching strategies (CDN/app/DB), configures invalidation (TTL/event-based), Redis/Memcached/Varnish, stampede prevention. Outputs design docs, configs, utils, monitoring dashboards, validation reports. Also supports assess, benchmark flags.

Audits an LLM prompt or pipeline for token inefficiency and produces an optimized version with caching configuration and estimated savings.

Audits a pasted prompt for structure, constraints, examples, context, edge cases, and versioning, producing a scored report with issues, recommendations, top improvements, and rewritten version if needed.

Applies 1h cache patch to Claude Code if needed, checks patch status, or restores original state. Supports 'patch' (default), 'status', 'restore' args.