The playbook series

How to use Claude well. From your first prompt to your first agent team.

An eight-part series for individuals upskilling on AI. Plain English, no code, field-tested in real work. Each playbook stands alone. Read in order, or pick the one that matches the problem in front of you.

Playbooks

Lines of code

Entry paths

Through-line

// Where to start

Three ways in.

Most readers won't read all eight in one sitting. Pick the path that matches where you are right now. You can always come back for the rest.

// New to AI

Start with the foundations.

Most "AI doesn't work for me" is a prompting problem, not a model problem. Start by learning to write a good prompt, then how to stop re-explaining yourself every chat. After that, the rest of the series clicks into place.

01 Prompting · the foundation
02 Memory & context · stop repeating yourself
03 Skills · your first reusable thing

Start with Playbook 01 →

// Already building

Build, connect, measure.

You've got a working setup and want to ship things that hold up. Start by making your work reusable, give Claude the data it needs, and build the discipline that separates real automation from theatre.

03 Skills · package what works
04 MCP · connect the tools
05 Evals · know if it works

Start with Playbook 03 →

// Leading a team

Spread it. Then scale it.

You've figured out something useful and want your team to use it too — without becoming the help desk yourself. Then take it further with single agents and then small agent teams that compress the work between decisions.

06 Rollout · spread without bottlenecking
07 Agent · one agent, end to end
08 Agent team · multiple agents working together

Start with Playbook 06 →

// The full series

Eight playbooks. One series.

Each playbook stands alone — but they build on each other. The order below is the natural learning order. Pick where to enter, then use the others as you need them.

// 01 — FOUNDATION

Six patterns

The language inside every Skill, agent, and instruction. The foundation everything else sits on.

~20 min read · 6 patterns

Open playbook →

// 02 — FOUNDATION

Memory and context

Four layers of what Claude knows. Set them up once. Watch every prompt shrink to 20 words with the same output.

~15 min read · 4 layers

Open playbook →

// 03 — BUILD

Build a Skill

Teach Claude how to do a recurring task. The first thing you'll package and reuse. Plain English, no code.

~15 min read · 20 min to build

Open playbook →

// 04 — REACH

Connect MCP

Give Claude access to the tools and data it needs. Calendar, email, drive — built around one real ritual.

~8 min read · 30 min setup

Open playbook →

// 05 — DISCIPLINE

Run evals

Know if it works. The cheapest piece of infrastructure you'll ever own. Trust through evidence, not optimism.

~½ day to build · 14 steps

Open playbook →

// 06 — SPREAD

Roll out to your team

Spread capability without becoming the help desk. Five names, public wins, the slow-and-then-quick rhythm of real adoption.

One agent. One job. Five build steps. The agent does the typing. You do the judging. No code anywhere.

~2 hrs to build · 5 steps

Open playbook →

// 08 — CAPSTONE

Build an agent team

Four design decisions that apply to every team you'll build after this one. The capstone of the series.

~4 hrs to build · 4 decisions

Open playbook →

// The thread running through all of it

AI compresses the work between your decisions.

Every playbook in this series argues the same thing, from a slightly different angle. The agent does the typing. The Skill does the structure. The Project does the remembering. The eval does the checking.

What stays with you is the part nobody else can do — deciding what's worth doing, judging whether the output is right, knowing where the guardrails belong. Read in order, the eight playbooks are a working stack. Read individually, each one stands on its own.

Either way, the bargain is the same: more time deciding, less time typing.

Looking for the rest of Kartini's work — case studies, expertise, contact?

Back to portfolio →