GitHub's trending page has been dominated by a single name recently: obra/superpowers, which rocketed to 270,000 stars within one week—reportedly peaking at 1,422 new stars in a single day, roughly the daily traffic of two mid-sized open source communities.
It is not a model, nor an IDE plugin, but a skill framework + software development methodology for AI coding agents. Author Jesse Vincent (obra) has packaged the engineering discipline the AI coding community has refined over the past six months—ask for requirements first, write plans, test-driven development, review before commit, verify before completion—into more than 20 composable skills.
How the Workflow Works
- When an agent starts and sees you want to build something, it doesn't immediately write code; it first steps back and asks what problem you're actually trying to solve.
- It refines requirements into a readable design document, confirmed with you in sections.
- Once the design is signed off, it breaks work into small tasks of 2–5 minutes each, with explicit file paths, code, and verification steps.
- It then enters subagent-driven development: each engineering task goes to a fresh subagent, followed by two-stage review (first checking spec compliance, then code quality), with critical issues blocking progress.
- Strict red-green TDD is enforced throughout: no implementation until tests pass, and failing code gets deleted and rewritten.
Cross-Tool Portability
The most underrated aspect is its cross-tool nature. It can be installed on 8 platforms: Claude Code's official plugin marketplace, Codex CLI, Codex App, Cursor, Gemini CLI, GitHub Copilot CLI, OpenCode, and even Kimi Code. Skills are written in Markdown + YAML frontmatter—human-readable, machine-parseable—with dependency management and verification hooks. This effectively extracts "engineering experience" from model weights into a portable, testable, auditable asset.
Industry Implications
The takeaway matters more than the star count. Over the past year, the AI coding narrative was a battle of model capability—Claude vs. GPT, whose code is better. Superpowers' popularity pivots the narrative: make agents reliable in engineering environments, rather than making models smarter. It even accelerates the commoditization of proprietary agent tools—portable skills mean Cursor and Codex can't lock users in, forcing competition on model quality and price alone.
This aligns with NVIDIA's recently disclosed AVO rein architecture: large companies proving "orchestration layer > model itself" within their own systems, while the open source community spreads the same philosophy across all tools via skill frameworks. Top-down and bottom-up, both point to the same judgment: AI coding's moat is migrating from model capability to the ability to "embed process into agent muscle memory."
Three Signals to Watch
1. Whether the .agents directory convention becomes a genuine cross-tool de facto standard.
2. Whether a paid skill marketplace emerges—and what a "skill economy" business model looks like.
3. Whether Anthropic and OpenAI break their own capabilities into portable skills—if they do, it means this layer is public infrastructure, not any single company's private moat.