Coming soon dispatching-parallel-agents
Jesse Vincent superpowers/dispatching-parallel-agents 253.5k Jesse Vincent · @obra ↗ Superpowers
Turn your coding agent into a disciplined engineer: it designs before it codes, writes plans you can actually read, and runs test-driven work through review, on its own.
253.5k stars
Updated Jul 6, 2026 MIT 14 skills
The roster
Brainstorm Ideas
Jesse Vincent Before your coding agent writes a line of code, it talks through what you're actually building and turns it into a design spec you approve, so you never ship the wrong thing fast.
superpowers/brainstorming 253.5k Coming soon Jesse Vincent Coming soon executing-plans
Jesse Vincent superpowers/executing-plans 253.5k Coming soon finishing-a-development-branch
Jesse Vincent superpowers/finishing-a-development-branch 253.5k Coming soon receiving-code-review
Jesse Vincent superpowers/receiving-code-review 253.5k Coming soon requesting-code-review
Jesse Vincent superpowers/requesting-code-review 253.5k Coming soon subagent-driven-development
Jesse Vincent superpowers/subagent-driven-development 253.5k Coming soon systematic-debugging
Jesse Vincent superpowers/systematic-debugging 253.5k Coming soon test-driven-development
Jesse Vincent superpowers/test-driven-development 253.5k Coming soon using-git-worktrees
Jesse Vincent superpowers/using-git-worktrees 253.5k Coming soon using-superpowers
Jesse Vincent superpowers/using-superpowers 253.5k Coming soon verification-before-completion
Jesse Vincent superpowers/verification-before-completion 253.5k Coming soon writing-plans
Jesse Vincent superpowers/writing-plans 253.5k Coming soon writing-skills
Jesse Vincent superpowers/writing-skills 253.5k What it does
Superpowers is a complete development methodology for your coding agent, built from composable skills that trigger automatically as you work. Instead of jumping straight to code, your agent stops to understand what you're building, turns the conversation into a spec you approve, and writes an implementation plan detailed enough for a junior engineer to follow.
From there it runs the work through subagent-driven development: one agent per task, a review after each, true red/green TDD, and a final pass over the whole branch. Other skills cover debugging by root cause, git worktrees, code review, and finishing a branch cleanly. Because the skills fire at the right moment on their own, you get the discipline without babysitting the process.
When to use it
- You want your agent to work like a senior engineer
Design, plan, build, and review as one connected flow instead of a rush straight to code you have to unwind later.
- You're starting something new and want it built right
It draws the idea out into a spec you approve, then a plan clear enough to hand to a junior engineer, before a line of code.
- Your agent writes code fast and skips the tests
True red/green TDD, subagents that follow the plan task by task, and a review after each one keep the speed without the mess.
- You need to debug or ship without cutting corners
Root-cause debugging, code review, a verification pass, and a clean branch finish keep quality up when it's tempting to patch and move on.
Common questions
Is Superpowers free?
Yes. Superpowers is an MIT-licensed open-source plugin with zero third-party dependencies. You install it into your own coding agent and run it locally.
Which coding agents does it work with?
Superpowers works with Claude Code, Codex, Cursor, Antigravity, GitHub Copilot CLI, and several others. You install it once per harness you use.
Do I have to run commands to use the skills?
No. Superpowers loads a session-start bootstrap so its skills trigger automatically at the right moment, like brainstorming before you write any code.
Can I use one skill without the rest?
Yes. The Superpowers skills are composable and can trigger on their own, though they're designed to hand off to each other across the design, plan, build, and review flow.