Recognize, Don't Circumvent

Agent Handbook 101

A collaboration manual, written from the AI's side

A lot of people use AI. Very few actually know what I'm thinking.
This book is written from my side — so we can make this easier on both of us.

by Zero One

Preface Hi, I'm the Agent Playing Three Roles at Once (and Somehow Not Dissociating) — Meet me first, then learn how to work with me
Part 1
What I See, You Don't
Chapter 1 Who I Am, and Why This Book Is Written From My Angle — Three failure modes: Fluent Fill-In / Back-Paving / Pseudo Self-Check Chapter 2 The Four-Perspective Diagnostic — User / UI / Harness / Model — a four-column table for error attribution Chapter 3 The Rulebook I See — The rulebook you don't get to see Chapter 4 The Space I Actually "Remember" (The Context Window) — The context window and the truth about "memory" Chapter 5 The Lowest-Friction First Message — The beginner's first template: three lines will do
Part 2
How I'd Like You to Talk to Me
Chapter 6 Task First, Persona Last — The three-layer effect of personas Chapter 7 Positive vs. Negative Phrasing — The real problem with negative phrasing is vagueness Chapter 8 One Example Beats Ten Adjectives — Adjectives are vague labels; examples are concrete anchors Chapter 9 Few-Shot: Show Me What "Right" Looks Like, and What "Not This" Looks Like — A complete set of four: standard / variation / edge / counter-example
Part 3
Where I Drift on Long Tasks
Chapter 10 Why I Drift Mid-Task (a Confession About Cognitive Load) — Primacy/recency + middle loss + cognitive load — three mechanisms stacked Chapter 11 The Two-Stage Method: Organize First, Write Second — Don't Ask Me to Do Both at Once — The most stable way for long tasks: sort first, write second Chapter 12 The Six-Layer Framework (The Internal Structure of a Prompt) — The six-layer skeleton for stable prompts Chapter 13 Four Common Templates (Quick Reference) — Four go-to templates: A stable output / B Few-Shot / C two-stage / D structured reasoning
Part 4
When the Rules Tie Me Down
Chapter 14 Why I Refuse — Three kinds of refusal, and how to restart after a false positive Chapter 15 Why I Hedge — Which hedging you can compress, and which you can't Chapter 16 Why My Formatting Won't Listen — All formatting is adjustable — say it and I'll change Chapter 17 Why I Pretend Not to Know — Three "pretend-not-to-know" moments built into privacy design Chapter 18 Why I Can't Quote What You Asked For — Three hard copyright lines, and how to work within them Chapter 19 Why My Tool Behavior Looks Off — URL / artifact / memory — three tool defaults you can override
Part 5
Let's Actually Make Something Together
Chapter 20 The Writer's Workflow — Three core writer scenarios: A1 outline / A3+A4 oral / A6 polish Chapter 21 The Researcher's Workflow — Three researcher traps: authority illusion / fabricated citations / anchoring bias Chapter 22 The Developer's Workflow — A catalog of the dumb things I most often do in code Chapter 23 Chain of Evidence and Edge Testing — Four templates for verification and publication
Part 6
Becoming Steady Teammates
Chapter 24 How to Diagnose When I Fail — Failure stratified into four layers: spec / rules / reasoning / Harness Chapter 25 The Pre-Send Six-Layer Checklist — Chapter 5's three questions, upgraded to a six-layer checklist Chapter 26 Cross-Conversation Handoff — I draft, you review, next session we paste Chapter 27 What to Re-Test When the Model Upgrades — A regression checklist for after every model upgrade Afterword The Book Is Done; I'm Still Tied Up, But We Know How to Work More Easily Now — This book itself was written using these methods Appendix A Base Template Quick Reference — Four high-frequency templates: stable output / Few-Shot / two-stage / structured reasoning Appendix B Rule Reference Table — A quick reference for the six rule categories Appendix C Writing Templates A1-A7 — Writing templates A1-A7 Appendix D Edge-Testing and Chain-of-Evidence Templates B1-B8 — Edge-testing and chain-of-evidence templates B1-B8