Configs2026年7月18日·1 分钟阅读

Khazix Skills — Curated AI Agent Skill Collection

A community-maintained collection of AI agent skills covering coding assistance, content creation, data analysis, and automation workflows. Ready to use with Claude Code, Codex, and other agent harnesses.

Agent 就绪

Agent 可直接安装

这个资产可安装;Agent 先选择当前运行时、检查安装计划,再运行匹配命令。

Native · 98/100策略:允许
Agent 入口
任意 MCP/CLI Agent
类型
Skill
安装
Single
信任
信任等级:Established
入口
Khazix Skills
直接安装命令
npx -y tokrepo@latest install e55c5cdb-82c5-11f1-9bc6-00163e2b0d79 --target codex

先 dry-run 确认安装计划,再运行此命令。

Introduction

Khazix Skills is a community-maintained repository of ready-to-use AI agent skill definitions. Each skill is a self-contained markdown file that gives an LLM agent specific capabilities, workflows, or domain expertise. The project aims to build a shared library so developers do not have to write every skill from scratch.

What Khazix Skills Does

  • Provides a categorized collection of agent skills covering coding, content creation, data analysis, and automation
  • Each skill file includes trigger conditions, step-by-step instructions, and output format specifications
  • Skills are designed to be portable across multiple agent platforms including Claude Code, Codex CLI, and Cursor
  • Includes a simple installation script for bulk-copying skills into your agent's configuration directory
  • Maintains a consistent skill format with metadata headers for easy discovery and filtering

Architecture Overview

The repository is organized as a flat collection of markdown files under a skills/ directory, grouped by category subdirectories (coding, content, data, automation, devops). Each skill follows a standardized template with frontmatter metadata (name, description, triggers, category) and a body containing the actual instructions the agent should follow. There is no runtime component — the skills are static text files consumed by whatever agent harness the user runs. A lightweight CLI script handles installation by copying selected skill files into the target agent's skill directory and validating the format.

Self-Hosting & Configuration

  • Clone the repository; no build step or dependencies required
  • Copy individual skill files or entire category directories into your agent's skill folder
  • For Claude Code, place skills in ~/.claude/skills/ or .claude/skills/ in your project root
  • For Cursor, adapt the skill content into .cursorrules or .cursor/rules/*.mdc format
  • Customize any skill by editing the markdown file directly — all instructions are plain text

Key Features

  • Growing library of production-tested skills contributed by the community
  • Standardized format makes skills interchangeable between different agent platforms
  • Each skill is self-contained with no cross-dependencies between skill files
  • Category-based organization helps find relevant skills quickly
  • Open contribution model with clear guidelines for submitting new skills

Comparison with Similar Tools

  • Awesome Claude Code Toolkit — aggregates external links to agents and plugins; Khazix Skills provides the actual skill files ready to install
  • Fabric (Daniel Miessler) — ships pre-built prompt patterns for specific tasks; Khazix Skills focuses on full agent skill definitions with triggers and multi-step workflows
  • Cursor Rules MDC Generator — auto-generates rules from documentation; Khazix Skills provides hand-crafted skills for specific use cases
  • Nuwa Skill — extracts skills from a person's content; Khazix Skills offers a pre-built library rather than a generation tool
  • OpenAI GPT Store — platform-specific custom GPTs; Khazix Skills are platform-agnostic markdown files

FAQ

Q: Which agent platforms are supported? A: The skills are markdown files that work natively with Claude Code. They can be adapted for Codex CLI, Cursor, and any agent harness that reads markdown-based instructions.

Q: Can I contribute my own skills? A: Yes. The repository accepts contributions via pull requests. Follow the skill template in the contributing guide to ensure format consistency.

Q: Do skills conflict with each other? A: Skills are independent files with their own trigger conditions. Conflicts are unlikely unless two skills have overlapping triggers, in which case the agent harness's priority rules apply.

Q: How are skills quality-controlled? A: Community review through pull requests. Maintainers check that skills follow the template format and that instructions are clear and testable.

Sources

讨论

登录后参与讨论。
还没有评论,来写第一条吧。

相关资产