Superpowers - Setup Guide¶
Source: obra/superpowers (1.2M+ combined installs) GitHub: obra/superpowers (261K+ ⭐) Category: Agent Infrastructure / Development Methodology Quality Tier: 🟢 Production
Superpowers is the most widely adopted agentic skills framework - a battle-tested development methodology that turns AI coding agents into systematic, reliable engineering partners. With 1.2M+ combined installs across 15+ skills, it covers the full software development lifecycle: brainstorming, planning, test-driven development, systematic debugging, code review, and more. For Hermes agents, Superpowers provides proven workflows that eliminate guesswork and enforce engineering rigor.
Installation¶
# Core workflow skills
npx skills add obra/superpowers --skill brainstorming
npx skills add obra/superpowers --skill writing-plans
npx skills add obra/superpowers --skill using-superpowers
# Development methodology
npx skills add obra/superpowers --skill test-driven-development
npx skills add obra/superpowers --skill systematic-debugging
# Collaboration + quality
npx skills add obra/superpowers --skill requesting-code-review
Included Skills¶
| Skill | Installs | Purpose |
|---|---|---|
| brainstorming | 294.5K | Structured ideation - explore solutions, evaluate trade-offs, converge on architecture |
| systematic-debugging | 199.6K | Root-cause analysis methodology - reproduce, isolate, fix, verify |
| writing-plans | 196.8K | Implementation planning - break down features into testable, sequential steps |
| using-superpowers | 194.9K | Framework orientation - how to combine and sequence Superpowers skills |
| requesting-code-review | 178.7K | Structured review requests - what to include, how to surface design decisions |
| test-driven-development | 177.1K | Red-green-refactor workflow - write tests first, implement, verify |
Prerequisites¶
| Requirement | Details |
|---|---|
| skills.sh CLI | npx skills (auto-installed) |
| AI coding agent | Claude Code, Codex CLI, or any skills.sh-compatible agent |
| Git | Required for code review and TDD workflows |
Key Capabilities¶
Brainstorming (294.5K installs)¶
The most-installed agent skill on skills.sh. Guides agents through structured exploration: define the problem, generate candidate solutions, evaluate trade-offs (performance, complexity, maintainability), and converge on an architecture decision. Replaces vague "what should I do?" agent loops with a decision framework.
Systematic Debugging (199.6K installs)¶
Eliminates debugging guesswork. Five-phase methodology: (1) reproduce the bug with minimal input, (2) isolate the failure to specific code paths, (3) root-cause analysis using git bisect or log tracing, (4) implement the fix, (5) verify with regression tests. Reduces debugging time by forcing methodological rigor instead of random print statements.
Test-Driven Development (177.1K installs)¶
True TDD workflow for AI agents: write failing test → implement minimum code to pass → refactor. Prevents over-engineering by enforcing "simplest thing that works" at each step. Includes patterns for mocking, fixtures, and test organization.
Writing Plans (196.8K installs)¶
Implementation planning that produces actionable steps - not vague descriptions. Each plan step includes: what to change, which files, what tests verify it, expected diff size. Allows agents to execute complex features without losing context mid-implementation.
Quick Start¶
# 1. Install the framework
npx skills add obra/superpowers --skill using-superpowers
npx skills add obra/superpowers --skill brainstorming
# 2. Start with a brainstorming session
claude "Using the brainstorming skill, explore architectures for a real-time collaboration feature"
# 3. Convert the decision into an implementation plan
npx skills add obra/superpowers --skill writing-plans
claude "Using the writing-plans skill, create an implementation plan from the brainstorming output"
# 4. Execute with TDD
npx skills add obra/superpowers --skill test-driven-development
claude "Using the TDD skill, implement step 1 of the plan"
Verification¶
# Check installed Superpowers
npx skills list | grep obra/superpowers
# Verify the framework loads
npx skills use obra/superpowers@using-superpowers
Notes¶
- Superpowers is framework-agnostic - works with Claude Code, Codex, and any skills.sh-compatible agent
- The
brainstormingskill alone has more installs than many entire skill catalogs - it's the de facto standard for agent ideation - Combine with
anthropics/claude-codeskills for the full agent development toolkit: Superpowers for methodology + Claude Code skills for extensibility - For Hermes agents: use
systematic-debuggingandwriting-plansbefore any implementation work - they enforce the rigor that prevents context-window sprawl - The
using-superpowersskill teaches the agent how to chain skills together (brainstorm → plan → TDD → review) for end-to-end feature delivery
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.