Superpowers 4.0 brings significant enhancements to its development workflow, particularly through subagent-driven development. The code review process has been refined into two distinct steps: a "spec review" and a "code review." The spec review ensures that the implementation aligns with the original plan, while the code review focuses on code quality. Both steps now function as formal loops, allowing for continuous feedback and improvement, which streamlines the development process. However, the integration of OpenAI Codex with subagents remains a challenge, delaying the removal of the previous workflow that required manual bridging. Changes to skill descriptions aim to improve clarity and effectiveness. The update targets how skills are triggered by focusing descriptions solely on when to use them, rather than what they do. For instance, the brainstorming skill now emphasizes its necessity before any creative work, leaving out extraneous details. Additionally, to address character limits and streamline functionality, several skills have been condensed into broader categories, like combining testing-anti-patterns into test-driven-development, which shifts the focus from what not to do to actionable guidance. The update also introduces a basic test suite, enhancing the reliability of the agent's workflow by verifying the full brainstorming, planning, and implementing process. Although not a comprehensive evaluation suite, these tests help ensure that skill triggering works as intended. Internally, Superpowers is adopting GraphViz's dot notation for process documentation, offering a clearer, less ambiguous way to represent processes. This shift may not be visible to users, but it aids in maintaining clarity within the system. The creator expresses a commitment to open-source development and has opened GitHub Sponsorships for those who wish to support the project.