superpowers

317.4K stars
MIT
Active this month
Last commit 11 days ago

Last verified Nov 14, 2026

Orchestrate coding agents through a skill-based workflow for planning, test-driven development, code review, and long-running implementation.

Visit website GitHub317.4K

GitHub preview for the Superpowers repository

Read the full breakdownWho it is for, how it works, trade-offs and the FAQ

who it's for

Who superpowers is for#

Developer using Claude Code

Installs Superpowers into Claude Code so all AI-assisted features start with a proper spec review instead of jumping directly to implementation.

Team enforcing consistent agent methodology

Standardizes on Superpowers to ensure every agent run from any developer's machine follows the same TDD methodology without manual CLAUDE.md maintenance.

Solo developer on a long project

Uses Superpowers to maintain discipline on a long solo project, letting the agent run autonomously for multi-hour sessions while trusting the spec-first guardrails.

Agency with client sign-off requirements

Installs Superpowers for client projects where the client must review and sign off on the implementation plan before any development begins.

Developer evaluating a new agent harness

Installs Superpowers first to validate whether the harness actually follows structured prompts before committing to a new coding agent.

the problem

The problem it solves#

Coding agents left to their own defaults write code before understanding the problem, skip tests, violate YAGNI, and drift from the original spec after a few turns. Custom CLAUDE.md files and system prompt patches help but are fragile, hard to maintain across projects, and do not enforce a consistent workflow from requirement gathering through implementation. There is no standard way to install a shared set of development standards that works across multiple agent harnesses.

how superpowers solves it

How it solves it#

Spec-first workflow with review gate

Spec-first workflow: the agent asks clarifying questions before writing any code and shows the spec in digestible chunks for review before implementation starts

Subagent-driven autonomous implementation

Subagent-driven implementation: after spec sign-off, the agent uses subagents to work through each engineering task, inspect results, and continue autonomously for hours without derailing

Enforces TDD, YAGNI, and DRY automatically

Enforces true red/green TDD, YAGNI, and DRY via skills that trigger automatically on recognized task patterns

Available on the Claude plugin marketplace

Available on the official Claude plugin marketplace: `/plugin install superpowers@claude-plugins-official`

Supports 8+ agent harnesses

Supports Claude Code, Codex CLI, Codex App, Cursor, Factory Droid, Gemini CLI, OpenCode, and GitHub Copilot CLI

MIT licensed, composable and forkable

MIT licensed; skills are composable markdown files you can fork and extend

strengths · trade-offs

Strengths and trade-offs#

Strengths

Automatic triggering, no per-project configSkills trigger automatically without any per-project configuration once installed at the harness level
One methodology across all your harnessesSupports multiple agent harnesses, so one methodology applies across your entire toolchain regardless of which agent you use on a given day
Spec review catches misunderstandings earlyThe spec-review loop catches misunderstandings before long implementation runs, not after hours of agent work that misses the actual requirement
MIT licensed with plugin marketplaceMIT licensed with active development and a plugin marketplace for extending with additional skill packs

Trade-offs

-Spec gate is intentionally constrainingImposes a structured spec-first workflow: teams that want agents to start coding immediately without a requirements conversation will find it constraining by design
-Effectiveness varies with model capabilityEffectiveness depends on the underlying agent following skill prompts accurately; lower-capability models may not adhere consistently to the TDD and YAGNI enforcement
-Separate install step per harnessInstalling across multiple harnesses requires a separate install step for each; there is no single cross-harness install command

versus alternatives

superpowers vs alternatives#

superpowers is an MIT-licensed open-source workflow layer for coding agents such as Claude Code, Codex, Cursor, Gemini CLI, OpenCode, and GitHub Copilot CLI.

vs Cursor: Cursor gives developers a proprietary AI editor with integrated chat, autocomplete, and agent workflows. superpowers does not replace that editor; it adds portable skills for brainstorming, planning, worktrees, test-driven development, subagent execution, and review, so teams can carry the same engineering process across Cursor and other agent harnesses.

vs Claude Code: Claude Code provides the proprietary agent runtime and model access. superpowers sits on top as a methodology plugin, with skills that force design review, implementation planning, red-green-refactor discipline, and code review checkpoints before an agent declares work complete.

vs managed coding agents: Managed agents such as Devin-style services aim to own more of the implementation loop for you. superpowers is better when you want your existing agent tools to follow a visible, repo-local process; managed agents are still better when you want a hosted service to handle orchestration, credentials, and long-running execution infrastructure.

tech stack · detected from GitHub

What it's built on#

Languages: JavaScriptPythonTypeScript

frequently asked

FAQ#

Will Superpowers slow down my coding sessions by adding approval steps?

Yes, intentionally. The spec step adds an upfront conversation before implementation begins. In practice, this typically saves more time than it costs by preventing long agent runs that miss the actual requirement and need to be restarted.

Does it work with harnesses other than Claude Code?

Yes. Superpowers supports Claude Code, Codex CLI, Codex App, Cursor, Factory Droid, Gemini CLI, OpenCode, and GitHub Copilot CLI. Each harness requires a separate install step.

Are the skills visible and editable?

Yes. Skills are composable markdown files. Fork the repo, modify the skills, and install your customized version by registering your fork as a marketplace with your harness.

also worth a look

Similar open-source tools#

Modern Web Guidance Src

InDeveloper Tools

Modern web platform skills for AI coding agents

6,087 GitHub stars
Licence Apache-2.0

Docmd

Alternative toMintlify

Markdown to website, search, and AI context files in one command.

2,565 GitHub stars
Licence MIT

Paradev

Alternative toConductor

Open-source macOS workspace for parallel CLI coding agents

16,295 GitHub stars
Licence MIT

Hallmark

InDesign & Creative

Design skill for coding agents with 48 themes, real photos and lint.

55,156 GitHub stars
Licence MIT

CodeBoarding

Alternative toCodeDashboard

Review code changes as architecture maps, not line diffs

2,680 GitHub stars
Licence MIT

Dyad

Alternative toLovable

Local, open-source AI app builder with your own API keys

22,298 GitHub stars
Licence Apache-2.0 with separate pro-directory license

by category

Open source tools in these categories#

642

Inbox Zero

Alternative toSuperhuman

AI email assistant that drafts replies and clears your inbox

13,025 GitHub stars
Licence AGPL-3.0
#Email

rakazo

Alternative toGrok Bot

AI teammates you own: your keys, your model, your machine.

8,291 GitHub stars
Licence Apache-2.0
#AI Agents

SnitchFeed

Live monitoring on LinkedIn, X, Reddit & more. Agentic social listening for startups.

Marketing & Growth

Advertise your product

The projects we add and every new guide. Free, unsubscribe any time.

Find open source alternatives to paid software

Who superpowers is for#