kiro-spec-tasks

To generate implementation tasks from requirements, you map approved requirement IDs and design artifacts to concrete 1-3 hour work units, producing an executable tasks.md file with defined boundaries and dependencies.

The best way to size tasks for autonomous implementation planning is to constrain each work unit to 1-3 hours, annotate architecture boundaries, and mark parallelizable tasks with explicit cross-boundary dependencies.

A task-graph sanity check works by running an independent review alongside a task-plan review gate to validate that every numeric requirement ID maps to at least one task with observable done criteria before updating spec.json.

Yes, you can merge newly generated tasks with an existing tasks.md file. The generation process merges new task definitions with existing content, writes the updated tasks.md, and updates spec.json metadata and approvals.

Yes, task generation supports auto-approval flags for unattended runs. This allows the task-plan review and task-graph sanity check to execute and update spec.json metadata without manual intervention during spec-driven development.

You need to map numeric requirement IDs to implementation tasks when preparing a feature spec for autonomous implementation, ensuring every requirement has corresponding tasks with observable done criteria and validated dependency graphs.