writing-plans

To create a test-first implementation plan, you decompose feature specs into bite-sized TDD tasks: write failing tests, run commands with expected outputs, implement minimal code, and commit. The plan includes exact file paths, full code snippets, and self-review checks for context-free execution.

Task decomposition for multi-step requirements breaks ambiguous specs into small, standalone actions mapped to exact files. It ensures an engineer with minimal repository context can implement features across files, tests, and CI by following explicit run commands and commit instructions.

Writing implementation plans for context-free engineers requires listing exact file paths, describing each file's responsibility, providing full test and code snippets, and specifying run commands with expected outputs. This ensures end-to-end execution without prior repository knowledge.

TDD planning works for multi-subsystem features by producing one or more standalone plans from a product spec. Each plan maps exact files, includes type-consistency checks, and results in working, testable software with a clear commit history.

Structuring engineering docs for subagent execution involves saving test-first plans to a specific path like .agents/superpowers/plans/YYYY-MM-DD-feature-name.md. The plan provides explicit execution options, placeholder scans, and self-review checklists for autonomous or inline processing.

Yes, implementation plans explicitly include commit commands and expected test outputs alongside full code snippets. This ensures precise execution validation, enabling engineers to verify minimal code implementations against failing tests before committing changes.