Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
This collection provides a structured set of skills for AI agents to manage the transition from initial concept to final implementation. It establishes a rigorous process for exploring user intent and design requirements before code is written.
The framework enables automation builders to standardize complex SDLC activities, such as dispatching parallel agents for independent tasks and executing multi-step implementation plans.
WHO IT'S FOR
AI agent / automation builders
implementing subagent-driven development workflows
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
In opencode.json, set the V2 plugins entry to {"plugins":["C:\\Users\\<you>\\.config\\opencode\\node_modules\\superpowers"]}. Replace the author-provided <you> placeholder with the Windows username so the value is an absolute path; do not use ~ because OpenCode does not expand it.
Restart OpenCode so it loads the configured plugin and registers its skills.
Use the systematic-debugging skill to investigate one failing test or unexpected behavior before proposing any fix. Inputs: the repository path, the exact failing test or reproduction steps, the full error message or stack trace, and the command that reproduces it. Read the complete failure evidence, identify the most likely root cause, and return a concise evidence-backed diagnosis plus the next smallest verification step.
Implement one small behavior with red-green-refactor
Uses test-driven-development
Repository path
Precise behavior requirement
Relevant test file or test directory
Project test command
Use the test-driven-development skill for one small feature or bug fix. Inputs: the repository path, a precise behavior requirement, the relevant test location, and the project test command. Write the smallest test first, run it and confirm it fails for the expected reason, add only enough production code to pass, rerun the test, and then perform only a small safe refactor if needed.
Use the verification-before-completion skill to check one proposed completion claim. Inputs: the repository path, the exact claim to verify, and the full command that would prove it. Run that complete command fresh, read its full output and exit status, and report either the verified claim with evidence or the actual failing status without implying success.
Claude
Codex
Cursor
Gemini CLI
Copilot
OpenCode
Hermes
Antigravity
Grok
Devin
Qwen Code
Kimi Code
Factory Droid