Superpowers is an agentic skills framework and software development methodology designed to enable coding agents to work more effectively through a structured, composable system of skills and workflows. Written primarily in Shell, the repository provides a complete methodology that agents can adopt to improve their development practices, with particular emphasis on test-driven development, systematic debugging, and collaborative workflows.
The framework operates through a carefully sequenced workflow that begins before any code is written. When an agent encounters a development task, Superpowers activates a brainstorming skill that asks clarifying questions to refine rough ideas, explores design alternatives, and presents the resulting specification in digestible chunks for user validation. Once a design is approved, the agent creates an isolated workspace using git worktrees on a new branch and establishes a clean test baseline. The system then generates detailed implementation plans that break work into bite-sized tasks of two to five minutes each, with exact file paths, complete code specifications, and verification steps included for each task.
During implementation, Superpowers enforces a red-green-refactor test-driven development cycle where failing tests must be written before any implementation code. The framework supports subagent-driven development, dispatching fresh subagents to handle individual tasks with two-stage review processes that check both specification compliance and code quality. This approach enables agents to work autonomously for extended periods without deviating from the established plan. The skills library encompasses testing, debugging with systematic root-cause analysis, collaboration features including code review workflows, and meta-skills for creating new skills and understanding the system itself.
The repository has grown to 246,179 stargazers and 21,830 forks as of July 2026, with recent activity showing steady growth of seven additional stars and three additional forks. Issue and pull request response latency shows a median of zero hours with a mean of 26.9 hours across 850 tracked items, indicating active maintenance. The most frequently used issue labels are bug with 308 occurrences, enhancement with 274, and claude-code with 169, reflecting the framework's primary integration point. The primary maintainer obra has logged 1,921 events, with secondary contributors arittr at 314 events and lligirlburg-wq at 19 events.
Superpowers integrates with multiple coding agent platforms including Claude Code, Antigravity, Codex App, Codex CLI, Cursor, Factory Droid, GitHub Copilot CLI, Kimi Code, OpenCode, and Pi. Each integration follows platform-specific installation procedures, with Claude Code offering installation through Anthropic's official marketplace. The framework emphasizes philosophy centered on test-driven development, systematic processes over ad-hoc approaches, complexity reduction through simplicity, and evidence-based verification before declaring success.
The repository is linked through overlapping contributors to significant-gravitas/autogpt, activepieces/activepieces, and quivrhq/quivr, indicating connections within the broader AI agent development ecosystem. Prime Radiant, the organization behind Superpowers, offers commercial support and managed spending options for enterprise users. The project maintains an active community through Discord, with contribution guidelines requiring work on the dev branch and adherence to the writing-skills methodology for any new or modified skills. The MIT License permits broad reuse and modification of the framework.
