jnmetacode/superpowers-zh
6 skills3.2k installs29.6k starsGitHub
Install
npx skills add https://github.com/jnmetacode/superpowers-zhSkills in this repo
1Chinese DocumentationChinese Documentation is an agent skill for solo builders and small teams publishing Chinese-language technical docs that read native—not machine-translated. It encodes spacing between Chinese and English or digits, full-width punctuation in Chinese sentences, bracket and quote conventions, and numeric style aligned with common open-source 中文排版 practice. Because it only runs on explicit /chinese-documentation invocation, it avoids hijacking unrelated coding tasks while giving you a consistent editorial pass on README sections, API 说明, release notes, and mixed-language SDK docs. Use during Build docs work before Ship or Launch when Chinese readers are a primary audience; pair with your content workflow rather than replacing technical accuracy review.565installs2Using SuperpowersUsing Superpowers (Chinese edition) is a journey-wide meta skill from the Superpowers stack that runs at the start of conversations to enforce how agents find and load skills. It states that even a 1% applicability threshold means you must invoke the platform skill tool—never Read skill files ad hoc—and that user project instructions outrank skill defaults when they conflict. The document maps Claude Code’s Skill tool to Copilot’s skill tool, Hermes skill_view progressive loading, and Gemini activate_skill, with subagent exceptions when you are only executing a delegated task. Solo builders install it so brainstorming, planning, TDD, and review skills are not bypassed by generic chat behavior. It is the front door to the rest of Superpowers: treat it as mandatory hygiene before Idea-through-Operate work unless you are explicitly a subagent on a narrow assignment.556installs3Verification Before CompletionVerification Before Completion (Superpowers ZH) is a journey-wide discipline skill: before you say fixed, passing, or ready, you run the command that could falsify that claim and quote what actually happened. Solo builders use it to stop agents—and themselves—from shipping narrative success when CI would disagree. The workflow defines a hard gate function, tables of insufficient versus sufficient evidence, and explicit anti-rationalizations for fatigue and “just this once.” It extends beyond unit tests to builds, linters, requirement checklists, delegated agent work (verify via VCS diff), and TDD red-green integrity. Canonical placement is Ship review before PR, but the same ritual applies after Validate prototypes, Build features, Launch checklist items, or Operate hotfixes whenever you would otherwise commit or announce victory without output.533installs4Receiving Code Reviewreceiving-code-review is a Chinese-localized Superpowers agent skill that teaches coding agents how to handle pull request and code review feedback with technical rigor instead of social fluff. Solo builders using Claude Code or similar tools install it so that when comments land—especially vague or sweeping lists—the agent pauses, restates requirements, validates against the real codebase, and only then implements fixes in a sensible order. It distinguishes feedback from a trusted collaborator versus an external reviewer, adds a YAGNI check when someone asks for “proper” abstractions, and documents when to push back with evidence. The skill does not replace requesting review or writing tests; it sits on the human side of the review loop after feedback exists. Prism places it on Ship review as the primary shelf, but the same discipline applies whenever you reconcile external suggestions during Build integrations or Operate iteration. Pair it with your team’s CLAUDE.md or house style so hard gates on tone and verification stay consistent.528installs5Systematic DebuggingSystematic Debugging is a journey-wide agent skill from the Chinese Superpowers collection that forces root-cause investigation before any repair. Solo builders hitting flaky tests, production defects, slow performance, build breaks, or integration mismatches often default to quick patches that hide deeper issues; this skill replaces that habit with a four-phase flow and a hard gate against proposing fixes too early. Phase one emphasizes reading errors completely, reproducing reliably, reviewing recent git and config changes, and adding diagnostic logging at component boundaries when several layers are involved. Later phases cover hypothesis-driven isolation, minimal fixes aimed at the true fault, and verification that the issue stays resolved. The guidance applies equally when you are rushed or when the bug looks trivial—both are called out as anti-patterns for skipping the process. Use it whenever an agent might otherwise shotgun edits across your codebase.518installs6BrainstormingBrainstorming (Superpowers 中文) is a journey-wide agent skill that forces a structured conversation before any creative or behavioral change in your codebase. Written for solo and indie builders using Claude Code, Cursor, or Codex with Chinese procedural copy, it mirrors the English Superpowers ritual: explore repo context, optionally attach visual references in a standalone message, ask one clarifying question at a time, surface multiple architectures with honest tradeoffs, and obtain explicit approval on each design section. The HARD-GATE is non-negotiable—even trivial-looking work must pass because unchecked assumptions waste the most time on “small” tasks. When the design stabilizes, the agent persists a dated spec under docs/superpowers/specs/, runs an inline consistency check, and only then invokes writing-plans for an implementation plan. Use it at Idea when choosing what to build, at Validate when locking scope, at Build before new UI or APIs, at Ship before risky refactors, and at Operate when changing production behavior.515installs