task-planning

Break approved work into 2-5 minute tasks by specifying exact file paths, line ranges, and concrete code examples. Each task must be completable independently with zero context, include a failing test, verification steps, and implementation details, enabling deterministic commits and zero-context agent execution.

Structure each task with precise file paths, a failing test that validates the requirement, explicit verification commands, and implementation code. This approach satisfies mandatory test scaffolding, ensures incremental progress, and produces auditable commits that agents can execute without external context.

Yes. Task planning works with zero-context agents by embedding all necessary information—file paths, line ranges, test cases, and verification steps—directly in the task description. This enables parallel execution, clear handoffs, and independent task completion without context sharing.

Use task planning after proposal approval when you need to convert approved requirements into executable work. It's designed for projects requiring deterministic implementation, mandatory testing, and auditable commits—not for exploration or proposal-phase planning.

Yes. Each task must include a failing test and explicit verification steps. Tests validate that the task satisfies functional requirements, and verification steps confirm completion, enabling agents to work independently and produce reliable incremental commits.

Include exact file paths, line numbers, and concrete code examples in each task. Scope each task to a single, focused change—backend or frontend—with its own test and verification. This granularity and explicitness makes tasks independently executable within the target timeframe.