AI Playbook
Prompts, skills, hooks and routines for your role, checked against the sources every Monday.
Pick your role
16 roles · 12 guides · 16 tools · updated every Monday, last update 13 Oct 2026
An agent works in a loop around a model. The model sits at the center and the building blocks orbit it, with data flowing between them: Prompting, Skills, Hooks, Routines, Subagents, MCP, Evals. Each block links to its guide.
Role playbooks
Each playbook starts with the tasks AI does well in that job, then gives you the prompts, a skill, a hook and a routine to run them, with the checks a person still owns.
Guides
How to hand work to an agent and still own the result: skills it loads when a task fits, hooks that check every action, routines on a schedule, subagents for parallel work and MCP for your tools.
Start here
Agentic ways of working · Beginner · 18 min read
How a team works with coding agents every day: give them the right context, constrain them with hooks, permissions and CI gates, verify every change, and roll the model out over 30 days without losing control of what ships.
Read the guideCraft
Seven levels, from a clear one-line ask to a prompt you test on real examples and improve with data. Each level adds one technique: examples, structure, reasoning steps, tools and retrieval, a system spec, then evaluation.
Step 1 of 7 · Clear ask
Summarize these 40 support tickets into the top 5 issues.
Hover or tap a step to see what it adds.
Craft
A prompt that looks better on three examples can be worse on the next hundred. Test changes the way engineers test code.
Thirty to a hundred inputs from real work, each with the answer you expect.
Code checks for format and facts, a model judge for tone and completeness.
CI runs the set on every change and fails the build when the score drops.
Tools
What teams run between their apps and the models: gateways, serving, observability, evals and the controls around agents. Each deep dive covers setup, production settings and when to pick something else.
Frameworks and SDKs for agents that plan, call tools and hand work back.
Typed decision models and gates that choose when automation acts and when a person does.
One endpoint in front of every model provider, with keys, budgets and fallbacks.
Run open models on your own GPUs with batching and caching.
Traces, token counts, cost and latency for every model call and agent step.
Test sets, scores and CI gates that show whether a change helped or hurt.
Tune prompts and programs against a metric instead of by feel.
Open specifications that let tools from different vendors read the same data.
Updates
Every Monday the playbook is checked against vendor changelogs and docs. Here is what moved.
Free membership
The top of every page stays open to everyone. Members also get the part below the line on every page. Across the playbook: