Chapter 1: Prerequisites
The skill is preparation. Knowing what must be in place — and what must stay outside AI’s reach — before the first prompt is typed.
Chapter 2: Token Efficiency
The skill is cost awareness. Reading a session for what it spends, not just for what it produces — and changing habits when the spend is wrong.
Chapter 3: The AI Agent
The skill is reading AI at the level of the machinery, not the chat window. What runs where. What each turn carries. Why one prompt becomes many. With that mechanic in view, the rest of the book lands as consequence, not magic.
Chapter 4: The Prompt
The skill is prompting on purpose. Naming the shape of conversation before typing it — and matching that shape to the work, instead of defaulting to command and hoping for the best.
Chapter 5: AI Rules
The skill is rule authorship. Writing instructions AI applies on every relevant turn — precise enough to land, short enough to live with, and disciplined enough to resist the urge to patch every misbehavior with another rule.
Chapter 6: Auditability
The skill is making the project’s history survive memory. The reader’s own words become the project’s source of truth — captured automatically, so nothing important depends on recall.
Chapter 7: Sessions and Memory
The skill is steering sessions instead of drifting through them. Reading a session for cost, attention, and rot. Knowing when to end one. Building project memory in files that outlive every fresh start.
Chapter 8: Glossary
The skill is negotiating language with AI. Recognizing when AI is reading a word a different way — and installing the small machinery that forces a shared vocabulary on every turn.
Chapter 9: Artifacts
The skill is separating sources of truth from derived outputs. Facts stay where AI can reason about them. Visuals stay where humans can review them. A clean rebuild propagates every fix.
Chapter 10: Trust, but Verify
The skill is validation. Planning the check before the prompt. Stacking cheap methods. Pointing each instrument at the failure mode it actually catches — and never trusting AI’s tone as a signal of correctness.
Chapter 11: Course Correction
The skill is course-correction without breaking the project. Turning the validation tools inward, on the reader’s own input. Updating facts that have drifted. Surfacing contradictions that already live inside the files, before they ship.
Chapter 12: Interactive Wiki
The skill is using the specification as a live, queryable knowledge base. Asking the project for what is needed — instead of re-reading it — and trusting the answer because every claim cites its source.