superpowers

To enforce spec-first software development, start with brainstorming and design approval before writing code, then proceed through TDD implementation, systematic debugging, and branch finishing. This structured workflow prevents ad hoc coding by requiring design validation upfront.

The best way to implement TDD with subagents is to create detailed execution plans and write failing tests first, then spawn dedicated subagents via sessions_spawn to handle implementation and code review. This subagent-based delivery ensures reliable outcomes through coordinated task execution.

Debug test failures through systematic root-cause analysis by diagnosing the underlying failure points rather than applying quick patches. This requires identifying the exact origin of the failure within the spec-first workflow before attempting any code modifications.

Yes, you can use this structured workflow for both bug fixing and feature development in agent-driven coding sessions. It applies spec-first rigor, TDD implementation, and branch completion processes to any software delivery task, ensuring consistent quality.

To complete a branch after tests pass, follow the defined closing process to merge, create a PR, keep, or discard your work. This branch finishing step verifies all tests and ensures the completed code meets the required guardrails before finalizing.