Superpowers

Opinionated agentic skills framework that turns coding agents into disciplined, spec-driven software engineers.

Freemium· Free (MIT); paid commercial support via PrimeradiantAgentsMulti-model7.1 / 10

Best for

Pick Superpowers if you run a coding agent for hours at a time and want it to behave like a senior engineer who plans, tests, and reviews before shipping.

Skip if

Skip it if you mostly want a chat-style assistant for ad-hoc snippets - the spec-and-plan ceremony will feel like overhead.

Superpowers is a software development methodology delivered as a composable set of skills and bootstrap instructions for coding agents. Instead of letting an agent leap into code, it forces a spec-first conversation, then produces a chunked, readable plan, then dispatches subagent-driven-development with strict TDD, YAGNI and DRY discipline, and reviews the work as it goes. The result is an agent that can run autonomously for hours without drifting from the agreed plan.

It is built by obra (Jesse Vincent of Primeradiant) and is harness-agnostic: Superpowers ships as an official plugin for Claude Code, Codex App and CLI, Cursor, Antigravity, Factory Droid, Gemini CLI, GitHub Copilot CLI, Kimi Code, OpenCode and Pi. The repository is MIT-licensed and free to install; Primeradiant offers paid commercial support, additional tooling and managed spend for enterprise users. It's aimed at engineers who already use coding agents heavily and want to impose process on them.

Under the hood it's a library of skills that auto-trigger based on context, plus bootstrap prompts that re-inject themselves after compaction, so you don't have to remember to invoke anything. Because skills are composable, teams can extend or replace pieces to fit their own SDLC, and the same install works across whichever harness you happen to be using that week.

Editor's take

This is the closest thing to a real SDLC for coding agents, and it's free. The subagent-driven-development loop genuinely keeps long-running runs on the rails, and the fact that it installs cleanly into nearly every major agent harness is rare. If you take agentic coding seriously, install it.

— The AI Tool Bible editorial team

Pros

✅ Works across Claude Code, Codex, Cursor, Gemini CLI, Copilot CLI and more from one repo
✅ Forces spec-first, TDD and subagent review instead of vibe-coding
✅ Skills auto-trigger after compaction so the discipline survives long sessions
✅ MIT-licensed and installable from official plugin marketplaces
✅ Backed by a commercial entity (Primeradiant) for enterprise support