scaffold

Scaffold: Build-Order Planning

Take a chosen design (from diverge-prototype or any architecture decision) and spawn N independent agents, each proposing a different build-order strategy. Each produces a sequenced implementation plan with milestones, dependencies, and a risk assessment; the results synthesize into one recommended build plan.

Arguments

[N] [path/to/design.md | inline description]

• N is the number of strategies (default 4, min 2, max 6).
• Input is a path to a design doc/prototype/PRD, or an inline description of what needs to be built. Missing or unclear: ask the user to clarify before starting.

Phase 1: Understand the Design

Read the design doc, prototype, or PRD provided. Extract the following:

• Components to build: every distinct module, service, or subsystem
• Dependencies between them: what requires what
• External integrations: third-party APIs, databases, services
• Testing requirements: what needs testing and at what level
• Deployment needs: infrastructure, environments, CI/CD

Prepare a build brief that summarizes:

• What needs to be built
• What exists already (if working within an existing codebase)
• What the constraints are (technical, timeline, team)

Present the build brief to the user and confirm before proceeding. Adjust if the user gives feedback.

Phase 2: Spawn Sequencing Agents

Launch all N agents in parallel using the Agent tool. Each agent receives the build brief plus a unique sequencing strategy drawn from this pool (assign strategies 1 through N):

1. "Riskiest-First": Start with the highest-uncertainty components. Front-load risk to learn early whether the approach works. Motto: "fail fast on the hard stuff."
2. "Demo-able First": Start with the components that produce visible, testable output. Build stakeholder confidence early. Motto: "something to show every sprint."
3. "Dependency-Topological": Strict dependency order. Build foundations first, layers on top. Nothing starts until its dependencies are complete. Motto: "no stubs, no mocks, each piece works when built."
4. "Vertical Slice": Build one thin end-to-end path through the entire system first. Proves integration works before widening. Motto: "narrow and deep before wide."
5. "Test Infrastructure First": Start with test harness, CI/CD, monitoring, observability. Build the ability to verify before building things to verify. Motto: "confidence before velocity."
6. "Parallel Tracks": Identify independent work streams that can proceed simultaneously. Optimize for team throughput. Motto: "max parallelism, defined interfaces."