Getting Started
Set up your own ACT Playbook workspace from the template and install the skills your team uses every day.
The ACT Playbook template is a ready-made repository you copy to create your own private workspace. It ships with a pre-configured CLAUDE.md, a memory file, and the folder structure your AI agent expects, so you can start working with the skills right away.
This guide takes you from an empty machine to a working playbook in three stages: create your repository from the template, install the ACT Playbook skills, and invite your team.
Create Your Repository
Copy the template into your own private repository on GitHub or any other platform.
Install Skills
Install the ACT Playbook skills and configure the API keys some of them need.
Collaborate
Invite your team and learn the day-to-day workflow.
What's in Your Workspace
Once you have your own copy, the repository is your private workspace. Here is what each part is for.
.agents/skills
Project-specific skills that only apply to this repository. For example, once you run the design system extractor, the resulting skill with your custom design tokens lives here. Skills in this folder are only active when you work inside this project.
context
Background information your AI agent uses to do better work. This includes competitive research, voice profiles for each team member, wording guidelines, and anything else that gives the agent context about your company and how you communicate.
workspace
Your actual working files. Website analysis, generated page content, sales campaign materials, social posts, and anything else produced during your day-to-day work goes here.
CLAUDE.md
The standing instructions your AI agent reads at the start of every conversation. The template ships with a pre-configured version that tells the agent where to find things in the repository, how to apply voice profiles, and a strict set of writing rules that keep output from sounding generic. This includes a banned vocabulary list (words like "leverage", "seamless", and "game-changer") and banned sentence patterns such as staccato rhythm, negation-contrast formulas, and throat-clearing openers. You can extend these rules with your own company style at any time.
MEMORY.md
A running log the agent keeps updated over time. When you correct it, share preferences, or give context about your team and projects, it writes the relevant details here. The next time you open a conversation, it reads this file first and picks up where you left off, without you having to repeat yourself.
User-Level Skills vs. Project-Level Skills
Some skills, like the website builder and its individual steps, are installed at the user level. They are available in any repository on your machine, not just this one.
This matters because your website might live in a completely separate repository. You can run the full website workflow there and tell the agent how to adapt: for example, "do not generate HTML files, write the content as Markdown so I can import it into our CMS" or "follow the file structure of this project when generating pages." The skill works the same way, you just give it the right instructions for your context.
Project-level skills, by contrast, are specific to one repository. Use them when a skill has been customized for your project, such as a design system skill that knows your exact color tokens and component names.