Skip to content

jnMetaCode/superpowers-zh

⭐ 2,419  ·  Shell  ·  GitHub Repo

This project solves the problem of AI coding tools lacking structured, systematic workflows and Chinese-language support. It stands out by providing a fully localized Chinese edition of the popular superpowers framework (159k+ stars), adding 6 original China-specific skills, and offering one-command installation across 17 AI coding tools, bridging the gap between Western tooling and Chinese developer workflows.

agent-skills agentic-coding ai-coding chinese claude-code code-review cursor gemini-cli

1-Sentence Summary

Localizes the 116k-star superpowers framework for Chinese developers, adding 6 original skills and one-command install across 17 AI coding tools.

🔥 Key Capabilities & USP

  • Complete Chinese Localization of 20 Skills: Translates all 14 upstream skills (brainstorming, TDD, debugging, code review, etc.) into Chinese while preserving English technical terms. This removes the language barrier that prevents Chinese-speaking teams from adopting structured AI workflows.
  • 6 China-Original Skills: Includes chinese-code-review, chinese-git-workflow, chinese-documentation, chinese-commit-conventions, mcp-builder, and workflow-runner. These are tailored for Chinese development ecosystems (Gitee, Coding) and team communication norms—something no Western fork provides.
  • One-Command Universal Installation: The npx superpowers-zh command auto-detects which AI coding tool your project uses (Claude Code, Cursor, Windsurf, etc.) and installs all 20 skills to the correct directory. This eliminates the manual, error-prone setup that plagues similar frameworks.
  • 17-Tool Support: Supports Claude Code, Copilot CLI, Cursor, Windsurf, Gemini CLI, Trae, Qwen Code, Kiro, and 9 others—nearly 3x the tool coverage of the upstream project. This future-proofs your team regardless of which AI assistant they adopt.
  • Systematic Workflow Enforcement: Forces AI agents to follow structured, battle-tested methodologies (brainstorm → plan → TDD → implement → review). This directly solves the "AI codes prematurely and hallucinates" problem plaguing ad-hoc prompting.

Technical Architecture

ComponentDescription
Packagenpm package (superpowers-zh) acting as installer + skill manager
Skill FilesMarkdown (.md) files containing structured, multi-step prompts for each workflow
Detection EngineAuto-detects project's AI tool (Claude Code, Cursor, etc.) by scanning config files
Installation MechanismCopies skill .md files into tool-specific directories (e.g., .claude/skills/)
Bootstrap FileCreates/updates CLAUDE.md (or equivalent) that references all installed skills
Uninstall--uninstall flag reverses all file operations cleanly
Upstream SyncMaintains a fork relationship with jnMetaCode/superpowers for periodic updates

The architecture is file-based and tool-agnostic: it leverages each AI coding tool's native skill-loading mechanism rather than building a proprietary runtime. This makes it lightweight, portable, and compatible with any tool that reads Markdown instructions from a project directory.

Quick Start Guide

bash
# Step 1: Navigate to your project
cd /your/project

# Step 2: One-command install (auto-detects your AI tool)
npx superpowers-zh

# Step 3: (Optional) Explicitly specify a tool
npx superpowers-zh --tool copilot

# To uninstall / clean up
cd /your/project
npx superpowers-zh@latest --uninstall

That's it. The installer detects whether you're using Claude Code, Cursor, Windsurf, Copilot CLI, or any of the 17 supported tools, then copies all 20 skill files to the correct location and creates the bootstrap file.

Pros, Cons & Use Cases

Pros

  • Dramatically reduces AI hallucination and premature coding by enforcing structured workflows (brainstorm → plan → TDD → implement → review)
  • One-command install across 17 tools—zero manual configuration
  • Actively maintained with upstream sync and WeChat community support
  • 6 China-original skills that address real gaps in Gitee/Coding workflows and Chinese team communication

Cons

  • 4 chinese-* skills require manual invocation (/chinese-xxx), not automatic—users must remember to trigger them
  • Fork dependency: may lag behind upstream superpowers updates if maintainer falls behind on sync
  • Overly prescriptive for simple tasks: the structured skills add overhead for trivial code changes where a quick prompt suffices

Who should NOT use this?

  • English-only teams who have no need for Chinese localization or Gitee/Coding workflows
  • Developers who prefer ad-hoc prompting and find structured workflows restrictive
  • Teams using unsupported AI coding tools (e.g., GitHub Copilot in VS Code without CLI, or custom in-house agents)

Ideal Use Cases

  • Chinese software teams adopting Claude Code, Cursor, or Copilot CLI who need structured workflows in their native language
  • Teams working with Gitee or Coding platforms that require Chinese commit conventions and code review standards
  • Organizations standardizing on AI coding tools who want a consistent, battle-tested workflow across their entire engineering team
  • Developers frustrated by AI hallucinations who want to force the agent to plan and test before writing production code

Community & Activity

With 2,419 stars and active maintenance (last updated May 2026), this project has strong momentum. It's a specialized fork of the 116k+ star superpowers project, but has carved out its own community by solving a real pain point: making AI coding tools work for Chinese-speaking developers. The maintainer actively syncs with upstream and runs WeChat-based community support, which is a strong signal of long-term viability. The 17-tool support (vs. upstream's 6) shows the project is innovating beyond its fork origin.

Project data from GitHub API, updated in real-time