A strategy plan sets the overall goal, approach, and quality-assurance mechanism for a software initiative. It gives later planning enough direction without deciding every implementation detail upfront. section 6.3 section 6.5
For a core messaging feature, the strategy calls for an end-to-end browser demonstration and uses behavior-driven development (BDD): write Gherkin scenarios first, then implement outside-in until full-stack tests pass. It also identifies decisions that need attention, such as messaging scope, edit and delete semantics, the UI data layer, and design consistency. section 6.3
Listing 6.2 Part of Claude’s strategy plan, which sets the overall context for the approach
This phase builds **core chat, end to end, demonstrable in a browser**: channels and membership, send/edit/delete, history pagination, and real-time delivery — plus the React UI to show it working. The specs are written first as Gherkin and drive the implementation outside-in; the scenarios run full-stack through Playwright against the Docker Compose stack, so "passing" and "demonstrable" are the same fact.
This strategy excerpt connects the feature goal to its test and demonstration approach. section 6.3
Use the strategy as input to a delivery plan that breaks the feature set into manageable pieces. An implementation plan then details how to build each piece. Rough phases or specific design suggestions in a strategy need not dictate the later plans; review them for direction rather than treating them as final instructions. section 6.3 section 6.5 section 6.7
Turn a strategy into reviewable slices
Save the strategy somewhere accessible to a later session. In plan mode, point Claude at it and ask for a delivery plan broken into chunks that start with Gherkin tests and lead to a build plan. Choose a slicing approach that keeps review manageable, and ask for clarification when a proposed technical choice is ambiguous. section 6.3 section 6.6
Skim-review the strategy and delivery plans, then scrutinize the BDD scenarios before building. Set an explicit stop: the agent presents the finished scenarios and waits for approval. Iterate until they describe the behavior you want; only an explicit go-ahead starts implementation. section 6.5 section 6.6


