A Playbook is a workspace for one project. Stages, prompts, best-practice files, and the artifacts the agent produces live in one place and run in sequence, for the life of the project. Discovery runs before Design, Design before Build. Nothing gets lost when you close the tab.
Each Playbook ships with best-practice files for naming, governance, and sequencing - read on every prompt, so the work follows a real methodology out of the box rather than a generic prompt chain. Attach your own frameworks, internal standards, and prior artifacts, and the agent works to those too.
Inventories, design specs, build-ready change lists - every output is saved to the project, not buried in a thread. Come back weeks later, re-run a step with new context, and the work picks up where you left it.
“What used to take two weeks of digging now takes just 20 minutes. Being able to get into a Flow and ask the AI agent what is happening in this Flow is actually invaluable.”
A Playbook runs the whole project in one line, with no handoffs between phases. The output of Discovery feeds Design; the output of Design becomes the input for Implementation; the output of Implementation hands off natively to Sweep’s Build environment. The same agent and the same live-org context carry all the way through - the agent doesn’t just describe what should happen, it does the work.
Playbooks are shared across the project team by default, so good methodology spreads. Admins, architects, developers, and project leads all work from the same live system context and the same plan. Mark a Playbook private when it’s a personal scratchpad.