Superpowers: Claude Agent Framework
Structured agent skills for parallel, methodical software development
What You Can Do
Superpowers provides Claude with a complete framework of specialized skills that teach AI agents how to approach software development systematically. Use it to brainstorm ideas into designs, execute detailed plans with checkpoints, dispatch parallel agents for independent tasks, and finish development branches with proper integration. The framework enforces best practices like design-first development, test-driven workflows, and comprehensive verification before completion.
Features
Turn ideas into fully-formed specifications through guided dialogue. Explores requirements, proposes 2-3 approaches with tradeoffs, presents designs for approval, and documents specs before implementation begins.
Delegate independent tasks to multiple agents working concurrently. Each agent gets focused scope, clear goal, and isolated context. Perfect for fixing multiple test failures or investigating unrelated subsystems simultaneously.
Execute detailed implementation plans with bite-sized steps, verification checkpoints, and progress tracking. Plans drive focused work without context drift or scope creep.
Automated integration decision tree that verifies tests, detects environment (local or worktree), presents merge options, and handles cleanup. Prevents integration mistakes and orphaned branches.
Technical evaluation framework for handling code review feedback. Verify before implementing, push back with reasoning when needed, distinguish unclear feedback from actionable guidance.
Structured approach to TDD that keeps tests and implementation in sync, validates all code paths, and prevents test-only cargo cult.
Methodical debugging process that gathers signals (logs, screenshots, stack traces), builds hypotheses, and tests them systematically before guessing.
Write reusable Claude skills that teach agents domain-specific capabilities. Skills compose into larger workflows through clear interfaces and dependencies.
Example Output
Brainstorming a new feature: Design before coding
Brainstorm: Add dark mode support to the application
Claude explores your codebase, asks clarifying questions about scope, proposes 2-3 different architectural approaches with tradeoffs, presents a detailed design covering components and styling, then commits an approved spec to docs/ before implementation. This prevents expensive rework from unclear requirements.
Fixing multiple test failures in parallel: Run independent investigations simultaneously
Agent 1 → Fix src/tests/auth.test.ts (3 failures)
Agent 2 → Fix src/tests/payment.test.ts (2 failures)
Agent 3 → Fix src/tests/dashboard.test.ts (1 failure)
Each agent independently investigates their test file, identifies root causes, and implements fixes. All three run concurrently instead of sequentially, cutting investigation time by 3x while preventing conflicts since each agent touches different code.
Executing a plan with verification: Step through implementation with safety checkpoints
Execute step 1: Create database schema (verify: schema migrates cleanly)
Execute step 2: Add API endpoints (verify: tests pass, logs clean)
Execute step 3: Add frontend components (verify: renders without errors)
Each task is small, well-defined, and verified independently. Claude reports progress and stops immediately if verification fails, preventing silent failures that propagate downstream.
What's Included
- 14 Specialized Skills: Complete skill library covering brainstorming, planning, execution, debugging, code review, testing, and branch management. Each skill is a reusable agent capability.
- Plan-Driven Framework: Structured methodology for turning ideas into designs, designs into plans, plans into code, and code into production. Includes decision trees and checklists for each phase.
- Parallel Agent Coordination: Framework for dispatching multiple agents to independent problems simultaneously. Includes isolation patterns, context management, and integration verification.
- Integration Verification: Automated workflows for verifying test suites, handling branch merging, creating PRs, and managing worktree cleanup. Prevents integration accidents.
- SKILL.md Templates and Examples: Documented skill structure with working examples. Teaches how to author new domain-specific skills that integrate seamlessly with the framework.
- Development Methodology: Complete SDLC philosophy covering design-first, test-driven, parallel development, and comprehensive verification. Works for teams of AI agents and humans working together.
Who It's For
- Software Development Teams
- AI Engineers Building Multi-Agent Systems
- Engineering Managers Coordinating Parallel Work
- Technical Architects Planning Complex Features
- Code Review Leads Establishing Development Standards
Best For
- Feature Planning and Design
- Parallel Development Coordination
- Systematic Debugging and Problem-Solving
- Test-Driven Development Workflows
- Code Integration and Branch Management
- Large Feature Decomposition
You might also like
The Lark CLI is the official command-line interface for Lark (飞书), packed with 35+ AI agent skills and 200+ commands. Automate workflows across your entire Lark workspace: manage approvals and attendance, develop custom applications with Miaoda, query and analyze multi-dimensional base tables, handle documents and sheets, schedule meetings, send messages, manage mail and tasks, and integrate with external APIs. Built for both humans and AI agents with comprehensive business domain coverage.
RigorPilot is a comprehensive skill suite for deep learning research, providing auditable repository analysis, README-first reproducibility workflows, bounded exploration with evidence-based ranking, and publication-ready scientific documentation. Each skill enforces rigor through documented assumptions, comparability tracking, and explicit decision checkpoints. Use these skills to understand ML repository internals, faithfully reproduce published results, systematically explore novel ideas with proper baselines, and generate standardized research artifacts.
Access 29 production-ready Claude agent skills that wrap RunComfy's AI models for image generation, video creation, audio synthesis, and avatar videos. Skills include music generation (ACE Step with inpaint/outpaint), AI avatars and lip-sync (OmniHuman, Seedance, HappyHorse), image editing (inpainting, outpainting, face-swap, relight), video generation and extension, and upscaling. All skills are CLI-driven via the runcomfy command, with built-in schema validation, error handling, and examples for each model.
Elevate your interface design from strategy to pixel-perfect execution. Shape layouts and user flows before coding, critique designs against established heuristics, audit for accessibility and performance gaps, and polish for production. Enhance with purposeful animation, strategic color, harmonious spacing, and delightful moments. Includes comprehensive guidance for websites, dashboards, apps, components, forms, onboarding, and design systems across web and native platforms.
by pbakausThis collection of skills improves code quality, architecture documentation, and team workflow coordination. Review Architecture Decision Records with fresh perspective and rewrite them for clarity, implement type-safe AST visitor patterns for robust code organization, automate Biome linting upgrades across monorepos, and generate well-structured GitHub pull requests integrated with Linear project tracking. Perfect for teams that care about maintainable code, clear decision documentation, and smooth development workflows.
Access Vercel's official collection of agent skills for building and deploying modern web applications. Optimize React component architecture using composition patterns to avoid boolean prop proliferation and build scalable component libraries. Deploy applications to Vercel with flexible workflows supporting git integration, CLI authentication, and multiple fallback methods. Apply React and Next.js performance guidelines from Vercel Engineering to write efficient code and improve Core Web Vitals.
Generate batches of proven viral hooks across named-creator frameworks (MrBeast, Hormozi, Brunson, Koe, Welsh, Bush, Galloway). Write and audit short-form scripts with platform-specific pacing. Optimize Instagram Reels for sends-per-reach using Trial Reels testing. Craft captions and CTAs that match video payoff. Diagnose why openings flop and rewrite weak hooks against anti-patterns. The skills encode what patterns tend to lift retention in the first 1 to 3 seconds, then repurpose across platforms.
Design and build animations using a proven decision sequence covering when to animate, which tools to use (CSS, WAAPI, Motion.dev), property selection, easing curves, and accessibility. Review existing animations and audit codebases for quality issues. Understand and name motion effects using a comprehensive vocabulary glossary. Learn from Emil Kowalski's opinionated motion design philosophy built on decades of interface engineering experience.