writing-plans

To create an implementation plan, translate ambiguous requirements into actionable, test-driven tasks with explicit file paths, concrete code blocks, and commit checkpoints. This approach ensures the implementer knows exactly what to change and how to test it.

Architecture-first decomposition is the process of mapping file structures and component responsibilities before proposing any code changes. This method ensures the implementation plan aligns with the existing system architecture and reduces guesswork.

Writing a test-driven implementation plan for agent workflows requires defining bite-sized tasks with exact file touchpoints and commit checkpoints. The plan includes a self-review checklist to verify spec coverage and eliminate placeholders before execution.

Yes, implementation plans are specifically designed for features spanning multiple files or subsystems where the implementer needs explicit guidance. They provide concrete code blocks and exact file paths to ensure safe, incremental execution.

The best way to eliminate placeholders in spec-to-plan generation is to use a self-review checklist that verifies spec coverage. This ensures the final plan contains exact file paths and concrete code blocks before the execution handoff.

Yes, the implementation plan includes a final execution handoff that offers subagent-driven or inline execution paths. This allows the generated plan to be routed to subagents for automated implementation after passing the self-review checklist.