planning-and-task-breakdown

To turn approved requirements into executable implementation tasks, you need to break work into thin vertical slices with explicit task boundaries, write sets, and named verifiers. This ensures agents don't replan blindly or ship ambiguous work.

Vertical slice task breakdown for multi-file changes shapes work into Goal, Epic, Work Package, and Atomic Task layers. This structure provides acceptance signals and named verifiers so parallelization decisions and multi-agent dispatch boundaries are explicit.

To sequence implementation tasks when evidence is partial, apply risk-first ordering to tackle unknowns early and use confidence gating. This prevents moving forward on unverified assumptions before sufficient evidence is gathered.

Handling rollback and cleanup tasks for migrations or compatibility layers requires explicit planning within the task sequence. You must include cleanup tasks for release-state changes and flags to ensure safe execution and verifiable rollback paths.

If approved scope or Goal IDs are missing, task breakdown stops and routes back instead of inventing planning-time IDs. This scope safety mechanism requires evidence-backed scope reading to ensure accurate dependency semantics.

Dependency semantics gate parallel implementation tasks by producing explicit dependency edges using a Dependency Semantics Contract. This ensures only consumed outputs block downstream tasks, enabling safe multi-agent dispatch and accurate parallelization.