Playbook outline

The topics the playbook is planned to cover, in the order they are planned. Titles and scope may change before publication.

Placeholder. The copy on this page is a draft outline. Final copy is still to be written and checked.

Planned chapters

The operating model. You direct, agents build: roles, the plan-build-review-test loop, and what "done" means.
Plan files. Why a written plan beats a long prompt, and how to keep it current.
Where your agents run. Web, desktop, cloud and terminal workspaces, and how to choose.
Choosing tools and model tiers. A dated snapshot of options and how to match tiers to tasks.
Memory, ledgers and decision logs. How to stop agents from forgetting what was decided.
Lanes and parallel agents. One agent per lane, merge order and collision recovery.
Dual review. Using two independent reviewers and a severity rubric.
Tests and acceptance checks first. Trusting output you did not read line by line.
Guard hooks and safety rails. Limits that hold even when an agent makes a mistake.
Cost gates. Budgets, thresholds and stop conditions.
Handoffs. Agent to agent and agent to human, without losing context.
Secrets, data and untrusted input. Keeping credentials and private data out of prompts, repos and logs.
Unattended runs. What is safe to leave running, and what is never unattended.
Starting from what you already have. Existing brand guidelines and apps as the starting point.
Deployment and running. Options, rollback first, and monitoring.
Platform reviews and API approvals. Preparing submissions for third-party review.

Not published. There is nothing to buy or download yet, and no results, rankings or time-saving figures are claimed here.