Claude skill Anatomy explains prompt caching with a printing-press animation
The 20K-star Claude skill Anatomy visualizes prompt caching as a printing press: 10 locked lines of type for the cached prompt, one new line per call, 14.3x cheaper using real Opus 5.5 prices.
I asked a Claude skill how prompt caching works. It built me a printing press. That's Anatomy, 20K stars on GitHub. Ask how something works and it invents a machine whose working is the answer: > 10 locked lines of type: the 50K-token prompt it caches > One new line per call: the only tokens at full price > First call $0.258, the same call from cache $0.018 14.3x cheaper, worked out from real Opus 5.5 prices. It even checks its own drawing: 0 loose line ends out of 378. What would you ask it to explain first? Install line and 9 more skills in the article.
This post quotes another post: x.com/0x_kern/status/2113939185350357347