Advanced Claude - 9 books

This is the tier where a prompt stops being a message and becomes a specification.

Build prompts from named blocks you can edit independently. Replace a page of rules with

three well-chosen examples — and learn the accidental pattern that is this technique's

main failure mode. Decide how hard Claude thinks, and discover that a well-architected

prompt often runs correctly at a much lower setting than you expected. Package a way of

working into a skill colleagues use without knowing it exists. Connect Claude to live

systems, with a properly serious chapter on prompt injection. Read the price sheet

properly.

It ends somewhere most material on this subject never reaches: a system you own, with an

evaluation set underneath it, so when somebody changes it you can tell whether they

improved it.

Every model ID, per-token price, cache multiplier, effort level and character limit was

verified against Anthropic's published documentation, and every number in the cost

example was computed in Python. A cost guide with invented arithmetic is worse than none.

Prompt Architecture · Examples That Steer · Thinking, Effort and Depth ·

Skills · Connectors and Live Data · Claude Code at the Desk · The Model Menu and the API

· Long Work: Chaining and Agents · Capstone: Your Own Claude System