Agent Skills that make coding agents design distinctive frontends with art direction, real photos and in-browser visual checks instead of generic AI landing pages.
98.4K6.7KJavaScriptMITUpdated 4 days ago
obra/superpowers
A software development methodology and composable skills library that guides coding agents through brainstorming, planning, TDD, and subagent-driven implementation.
- Stars
- 301,139
- Forks
- 26,883
- Open issues
- 377
- Last commit
- 9 days ago
What is it?
Superpowers is a software development methodology for coding agents, built as a set of composable skills plus instructions that make the agent use them. When it detects that you are starting to build something, it pauses to clarify intent, presents the design in digestible sections, and then produces an implementation plan with exact file paths, code, and verification steps for each small task.
During implementation, the workflow emphasizes red/green test-driven development, YAGNI, and DRY. Work can be executed by dispatching a fresh subagent per task with review after each step, or inline in one session with a final review. Skills such as systematic debugging, code review requests, and branch completion are triggered automatically as the work progresses.
The project installs as a plugin for many coding harnesses, including Claude Code, Codex, Cursor, Gemini CLI, OpenCode, Pi, and others, each with its own install steps. It is MIT licensed.
Key features
- 01Turns a rough idea into a design through guided questions before any code is written.
- 02Produces step-by-step implementation plans with exact file paths and verification steps.
- 03Enforces red/green TDD cycles during implementation.
- 04Runs subagent-driven development with review between tasks, or inline execution with a final review.
- 05Includes skills for systematic debugging, code review, git worktrees, and branch completion.
- 06Installs as a plugin for multiple coding agents.
Use cases
- Having a coding agent build a feature from a vague idea without skipping design and planning.
- Running long autonomous implementation sessions that follow a written plan with review checkpoints.
- Standardizing TDD and code review practices across a team using multiple coding agents.
- Debugging a session where an agent ignored its plan or misused skills.
Topics
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source Python web crawler that turns any website into clean, LLM-ready Markdown for RAG, AI agents and data pipelines.
85.7K8.9KPythonApache-2.0Updated 3 days ago
A toolkit that lets AI coding agents clone a viral video into an editable, re-runnable video workflow with swappable footage, captions, and B-roll.
23K2.7KTypeScriptOtherUpdated 3 days ago
A macOS menu bar app and local gateway that lets Claude Code, Codex, Gemini CLI and other agents use any model or subscription you have.
15K1.1KGoMITUpdated 22 hours ago
A Claude Code and Codex agent skill that builds cinematic product promo videos in Remotion from your own screenshots and copy.
12.3K1.1KTypeScriptApache-2.0Updated 23 hours ago
Structured, verifiable AI-DLC software-delivery workflows that run as steering rules inside AI coding assistants like Claude Code, Kiro, Codex, and Cursor.
5.4K966TypeScriptMIT-0Updated 23 hours ago