superpowers-zh is the Chinese-language community edition of the superpowers agent-skills framework, packaging 20 skills — 14 translated from upstream plus 4 written for Chinese development workflows — for use across 26 AI coding tools, and it is aimed at Chinese-speaking developers and teams who already drive assistants such as Claude Code, Cursor, Windsurf, or Gemini CLI and want those assistants to follow a disciplined engineering method instead of jumping straight into code.

What it is

superpowers-zh is a JavaScript project distributed through npm as the package superpowers-zh, licensed under MIT, and developed in the open at github.com/jnMetaCode/superpowers-zh. It is a community translation and extension of obra/superpowers, the AI coding skills framework with more than 250,000 stars, and it keeps a public registry footprint under Developer Tools / Version Control & Collaboration with 8,146 stars, 758 forks, and 12 open issues. The project ships its documentation in Simplified Chinese (README.md), Traditional Chinese (README.zh-Hant.md), and a link to the English upstream README, with a project homepage at sp.aiolaola.com. It lives squarely in the npm and agent-skills ecosystem, and its topic list — agent-skills, agentic-coding, ai-coding, claude-code, code-review, mcp, npm-package, prompt-engineering — describes its territory precisely.

The concrete problem it solves is that AI coding assistants begin producing code before they understand the request. The README illustrates this with a batch-export feature: without the skills installed, the assistant immediately writes an exportUsers function that ignores the required output format, omits pagination, and risks running out of memory on large datasets. With superpowers-zh installed, the assistant first asks the clarifying questions that a careful engineer would ask. The project replaces prompt improvisation with reusable, versioned skill definitions that a team can install once and apply consistently, and it replaces English-only skill text for Chinese-speaking developers.

Key capabilities

• Installs from npm as superpowers-zh and targets 26 AI coding tools in one package.
• Ships 20 skills: 14 translated from upstream superpowers, 4 China-original skills, and 2 upstream skills retained for historical reasons.
• Supports Claude Code, Copilot CLI, Hermes Agent, Cursor, Windsurf, Kiro, Gemini CLI, Codex, Aider, Trae, VS Code (Copilot), DeerFlow, OpenCode, OpenClaw, Qwen Code, Antigravity, Claw Code, Qoder, CodeBuddy (Tencent), CodeArts (Huawei Cloud), Cline, Kilo Code, and Crush.
• Covers a full development methodology from brainstorming and planning through TDD, debugging, and code review.
• Provides parallel Simplified Chinese, Traditional Chinese, and English documentation paths.
• Exposes skills through MCP-oriented integration and publishes project documentation at sp.aiolaola.com.
• Bundles free companion learning material: the 182-lesson course "从零学会 AI 编程" and the 40-lesson "从零构建 AI 智能体", part of 13 courses and 648 lessons offered at ¥0.

Who uses it and how

• Individual Chinese-speaking developers who want Claude Code, Cursor, Windsurf, Kiro, or Gemini CLI to follow a repeatable workflow rather than improvise each task.
• Teams working inside Chinese enterprise toolchains such as CodeBuddy (Tencent), CodeArts (Huawei Cloud), and Qoder, where English-only skill text is a friction point.
• Terminal-first and agent users running Codex, Aider, OpenCode, OpenClaw, Cline, Kilo Code, or Crush who need the same methodology across several tools.
• Engineering groups standardizing TDD and code review practice so that assistants produce reviewable changes instead of unreviewed dumps.
• Learners pairing the installed skills with the free course track to move from prompting into structured agentic development.

Getting started

Install the npm package superpowers-zh, then point the target AI coding tool at the installed skills. Documentation and setup guidance are published at sp.aiolaola.com, with the package listing at npmjs.com/package/superpowers-zh.

How it compares

The direct counterpart is upstream superpowers from obra/superpowers, the original English framework with more than 250,000 stars, from which this project takes 14 translated skills and 2 retained historical skills. superpowers-zh is not a fork in competition with it: it is the Chinese community edition that tracks the upstream methodology while adding 4 China-original skills, so teams working primarily in Chinese get localized skill text and Chinese tool coverage, while teams working in English are better served by the upstream source directly. Both are MIT licensed.

When to use it — and when not to

Use it when the team works in Chinese, adopts several of the 26 supported tools, and wants a shared engineering method rather than per-developer prompting habits. Do not pick it if the team works in English — the upstream project is the original source and will always be ahead of a translation layer. Be aware of two practical frictions: the project depends on upstream's structure and naming, so translated skills track someone else's design decisions, and the README is heavily occupied by sponsor placements, which makes the setup path harder to scan than a lean documentation file would be. With 12 open issues and no server-side component mentioned, there is nothing to self-host, which lowers operational burden but also means the hosted documentation site is the practical reference point.

你：给用户模块加个批量导出功能
AI：好的，我来实现...（直接开始写代码）
    export async function exportUsers() { ... }
你：等等，格式不对，没分页，大数据量会 OOM...

你：给用户模块加个批量导出功能
AI：在开始实现之前，我需要了解几个关键问题：
    1. 导出格式是 CSV 还是 Excel？
    2. 预计数据量多大？需要异步处理吗？
    3. 有权限要求吗？
    → 给出 2-3 个方案，确认后再动手