Agentic Engineering Transition Playbook

You built a career writing code. Agentic systems are changing what that career looks like.

This is both parts of the Transition Playbook in one volume, 139 pages. Part 1 builds the foundation: the mental model of intent over implementation, breaking features into agent-sized chunks, writing CLAUDE.md as a contract with hard constraints instead of polite suggestions, and a review protocol for reading what the agent hands back. Part 2 goes further into orchestration: a five-level model for how much autonomy to grant, multi-agent chains and parallel workers, structured post-mortems when an agent fails, and integrating agent work into an existing Git workflow.

Both parts carry a 30-day transition plan, deliberately. Part 1 gives you the daily habits from a standing start. Part 2 rebuilds the same 30 days for someone already running agents in production. If you are starting from zero, read them in order. If you already ship with agents daily, Part 2 stands on its own.

What is inside:

* The agentic mental model: specifying intent instead of writing implementation

* Task decomposition that fits what an agent can actually hold

* CLAUDE.md patterns: hard constraints versus soft guidance, and why the difference matters

* A code review protocol for agent output

* Trust calibration: when to let it run and when to checkpoint

* Non-interactive mode: agents in CI and pre-commit hooks

* Debugging agent-generated code, and structured post-mortems for agent failures

* Multi-agent patterns: chains, parallel workers, and where they break

* Metrics that tell the truth about your own transition

* Common failure modes, with recovery paths

This is a field guide, not a survey of the literature. It is written for working developers who need to change how they work this quarter, not eventually.