cache

What problem does it solve? Claude Code bills are hard to interpret because cache reads are billed at a tenth of the input rate and usually dominate token volume, making raw token totals misleading. This Skill answers "why is my bill like this" with deterministic numbers from the metered waybill ledger instead of estimates. ## Core Features & Use Cases - Cache tier breakdown: Reports token volume by class — input, output, cache reads, and cache writes split by 5-minute and 1-hour premiums — plus the cache-read share of total volume. - Derived cost view: Computes effective cost, list-price equivalent, net caching savings, and coverage percentage from the current rate table, always labeled as derived rather than invoiced. - Per-model attribution: Breaks down volume and effective cost per model so you can see which model drives the bill. - Use Case: A budget owner sees an alarming token count and asks what caching actually saved; the Skill returns the net savings figure with its derived-basis label and the honesty floor of unattributed percentage. ## Quick Start Ask "why is my Claude bill like this" or "how much of my spend is cache reads" to get a receipt-backed breakdown of your token spend by cache tier.