← 返回
未分类

Superpowers

A complete software development methodology and skill composition system, ported from obra/superpowers. Provides structured, repeatable workflow: brainstorming, design, planning, TDD, subagent-driven dev, code review, finishing branch. Use for coding, building projects, refactoring, bug fixing, or any programming tasks.
A complete software development methodology and skill composition system, ported from obra/superpowers. Provides structured, repeatable workflow: brainstorming, design, planning, TDD, subagent-driven dev, code review, finishing branch. Use for coding, building projects, refactoring, bug fixing, or any programming tasks.
miketang
未分类 community v1.0.1 2 版本 99741.6 Key: 无需
★ 3
Stars
📥 326
下载
💾 0
安装
2
版本
#latest

概述

Superpowers

Overview

Superpowers is a complete software development methodology that provides AI agents with a structured, repeatable workflow. Ported from the obra/superpowers project, it breaks down the software development process into 7 core phases, each with dedicated skills and checkpoints to ensure high-quality, maintainable code output.

Core Workflow

Superpowers follows a strict 7-phase workflow. Each phase is mandatory and cannot be skipped:

1. Brainstorming

Trigger: Automatically activates at the start of any development task.

Goal: Clarify the user's real needs through questioning, explore alternatives, and produce a verifiable design document.

Key Actions:

  • Ask what the user really wants to achieve (not just surface requirements)
  • Explore multiple design approaches and trade-offs
  • Break design into readable chunks (max 200 words each)
  • Generate design document and save to design.md
  • Get user sign-off on the design

2. Design Validation

Trigger: After Brainstorming completes.

Goal: Ensure the design is complete, feasible, and meets user expectations.

Key Actions:

  • Present design chunks for user review
  • Ask "Is this section clear?"
  • Confirm technical feasibility and constraints
  • Record feedback and update design

3. Implementation Planning

Trigger: After design validation passes.

Goal: Decompose the design into executable, fine-grained development tasks.

Key Actions:

  • Create detailed task list, each task taking 2-5 minutes
  • Each task includes: exact file paths, complete code, verification steps
  • Follow YAGNI (You Aren't Gonna Need It) and DRY (Don't Repeat Yourself)
  • Generate plan document and save to plan.md

4. Test-Driven Development (TDD)

Trigger: Automatically activates during implementation.

Goal: Enforce the Red-Green-Refactor cycle.

Key Actions:

  • Write a failing test first (Red)
  • Write minimal code to pass the test (Green)
  • Refactor to keep code clean
  • Delete any code written before tests
  • Follow this cycle for every feature

5. Subagent-Driven Development

Trigger: After plan approval.

Goal: Use subagents to execute tasks in parallel or sequentially.

Key Actions:

  • Create fresh subagent per task
  • Two-stage review: spec compliance, then code quality
  • Set human checkpoints during batch execution
  • Stop and report immediately if subagent deviates from plan

6. Code Review

Trigger: Between tasks or upon completion.

Goal: Ensure code meets plan and quality standards.

Key Actions:

  • Review code against the plan
  • Report issues by severity: Critical (blocking), Major, Minor
  • Critical issues must be fixed before proceeding
  • Generate review report and save to review.md

7. Finishing Development Branch

Trigger: After all tasks complete.

Goal: Cleanly finish the development cycle.

Key Actions:

  • Verify all tests pass
  • Present options: merge, create PR, keep branch, discard
  • Clean up workspace
  • Generate final report

Core Principles

1. Test-Driven Development

  • Always write tests first
  • Red-Green-Refactor cycle
  • Tests serve as documentation and design tools

2. Systematic over Ad-Hoc

  • Follow process rather than guessing
  • Every decision has a rationale
  • Document all assumptions and decisions

3. Complexity Reduction

  • Simplicity is the primary goal
  • Remove unnecessary code
  • Prefer the simplest viable solution

4. Evidence over Claims

  • Verify rather than declare success
  • Use tests to prove functionality
  • Use data to support decisions

Skill Trigger Mechanism

Superpowers is a mandatory workflow, not an optional suggestion. When any of the following is detected, the corresponding skill must activate:

  1. User mentions development/programming -> Start Brainstorming
  2. Design complete -> Start Implementation Planning
  3. Starting to write code -> Start TDD
  4. Tasks need parallelism -> Start Subagent-Driven Development
  5. Code complete -> Start Code Review
  6. Project complete -> Start Finishing

Deliverables

Each phase produces documentation to ensure traceability:

PhaseOutput FilePurpose
----------------------------
Brainstormingdesign.mdComplete design document
Planningplan.mdDetailed implementation plan
Developmentcode/Source code
Reviewreview.mdCode review report
Finishfinal_report.mdProject summary

Example Workflow

User Request: "Build me a todo app"

  1. Brainstorming: Ask about requirements (Web/Mobile? Auth needed? Data storage?)
  2. Design: Generate UI design, architecture diagram, API design
  3. Planning: Break down into: create project structure, implement auth, build CRUD, add frontend
  4. TDD: Write tests for each feature first
  5. Subagent: Dispatch subagents to develop different modules in parallel
  6. Review: Review code quality of each module
  7. Finish: Merge code, generate deployment docs

Troubleshooting

Issue: Subagent deviates from plan

Solution: Stop immediately, analyze the deviation, adjust plan or re-dispatch

Issue: Tests failing

Solution: Return to Red phase, analyze failure cause, fix tests or code

Issue: Requirements change

Solution: Return to Brainstorming phase, re-evaluate design

Reference Resources

For detailed workflow guides and best practices, see references/workflow.md and references/best_practices.md.

Resources

This skill includes bundled resource directories for organizing different types of content:

scripts/

Executable code (Python/Bash/etc.) that can be run directly to perform specific operations.

Examples from other skills:

  • PDF skill: fill_fillable_fields.py, extract_form_field_info.py - utilities for PDF manipulation
  • DOCX skill: document.py, utilities.py - Python modules for document processing

Appropriate for: Python scripts, shell scripts, or any executable code that performs automation, data processing, or specific operations.

references/

Documentation and reference material intended to be loaded into context to inform the agent's process and thinking.

Examples from other skills:

  • Product management: communication.md, context_building.md - detailed workflow guides
  • BigQuery: API reference documentation and query examples
  • Finance: Schema documentation, company policies

Appropriate for: In-depth documentation, API references, database schemas, comprehensive guides, or any detailed information to reference while working.

assets/

Files not intended to be loaded into context, but rather used within the output produced.

Examples from other skills:

  • Brand styling: PowerPoint template files (.pptx), logo files
  • Frontend builder: HTML/React boilerplate project directories
  • Typography: Font files (.ttf, .woff2)

Appropriate for: Templates, boilerplate code, document templates, images, icons, fonts, or any files meant to be copied or used in the final output.


Any unneeded directories can be deleted. Not every skill requires all three types of resources.

版本历史

共 2 个版本

  • v1.0.1 Initial release 当前
    2026-05-26 01:48 安全 安全
  • v1.0.0 Initial release
    2026-05-26 01:34 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

ai-agent

Self-Improving + Proactive Agent

ivangdavila
自我反思+自我批评+自我学习+自组织记忆。智能体评估自身工作、发现错误并持续改进。
★ 1,387 📥 321,316
ai-agent

self-improving agent

pskoett
捕获经验教训、错误及修正内容,以实现持续改进。适用于以下场景:(1)命令或操作意外失败;(2)用户纠正Claude(如“不,那不对……”“实际上……”);(3)用户请求的功能不存在;(4)外部API或工具出现故障;(5)Claude发现自身
★ 4,090 📥 817,792
ai-agent

Skill Vetter

spclaudehome
AI智能体技能安全预审工具。安装ClawdHub、GitHub等来源技能前,检查风险信号、权限范围及可疑模式。
★ 1,233 📥 268,632