ouyangyipeng

Skill Compiler — AI skill for Claude Code

AI community

Write Once, Run Anywhere for AI Agent Skills.

How to install Skill Compiler

This entry records only its repository, not the path inside it, so there is no exact command to give. Open ouyangyipeng/Skill-Compiler and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Skill Compiler does

Write Once, Run Anywhere for AI Agent Skills. Claude Code. Codex. OpenClaw. Gemini.

Alternatives in AI

  • AI Guide — 程序员鱼皮的 AI 资源大全 + Vibe Coding 零基础教程,分享 OpenClaw 保姆级教程、大模型玩法(DeepSeek / GPT / Gemini / Claude)、最新 AI 资讯、Prompt 提 11.8k ★
  • Graphjin — GraphJin - An Automagical Compiler to Connect AI to Your Databases 3034 44 1 3k ★
  • Skillport — Bring Agent Skills to Any AI Agent and Coding Agent — via CLI or MCP 408 ★

README

SkCC Logo

SkCC

Compilation for Skills: Capable, Portable, and Securable

License Rust Platforms Paper

🌐 **Website**: [skcc.nexa-lang.com](https://skcc.nexa-lang.com) · 📄 **arXiv**: [2605.03353](https://arxiv.org/abs/2605.03353)

📚 **文档**: [中文](docs/USER_GUIDE.md) | [API Reference](docs/API_REFERENCE.md)


📦 安装

# npm(推荐 Node.js 用户)
npm install -g nexa-skill-compiler

# cargo(推荐 Rust 用户)
cargo install nexa-skill-cli

# 从源码编译
git clone https://github.com/Nexa-Language/Skill-Compiler.git
cd Skill-Compiler
cargo install --path nexa-skill-cli

🚀 快速开始

# 为所有平台编译技能
nsc build skill.md

# 为指定平台编译
nsc build skill.md --target claude

# 验证技能文件
nsc validate skill.md

# 从模板初始化新技能
nsc init my-skill

⚡ 什么是 SkCC?

**SkCC**(Skill Compiler for Cross-framework LLM Agents)是一个将经典编译器设计引入 Agent 技能开发的编译框架。通过四阶段流水线——Syntax Parser → IR Builder → Security Optimizer → Target Emitter——它将统一的 `SKILL.md` 源文件编译为面向 Claude Code、OpenAI Codex、Google Gemini CLI 和 Kimi CLI 的平台原生技能产物。核心创新在于 **SkIR**(统一中间表示),将技能语义与框架格式解耦,将适配复杂度从 $O(m \times n)$ 降至 $O(m + n)$,同时通过编译期 Anti-Skill Injection 实现安全加固。论文已被 **ACM CAIS 2026 — AgentSkills'26 Workshop** 接收(poster)。


📊 实验数据

所有实验基于 [SkillsBench](https://arxiv.org/abs/2602.12670)(89 个真实编程与数据分析任务),覆盖四个主流 Agent 平台。

EX1:编译增益 — 四模型对比

条件 任务数 通过 通过率 平均奖励
Claude-Baseline 38 8 21.1% 0.245
Claude-SkCC **