What it does

This skill enables developers to break the cycle of agent lock-in by refactoring AI instructions into a "Portable Core" model. It analyzes existing skills written for specific platforms—like Claude Code, Cursor, or GitHub Copilot—and decouples the business logic from the platform-specific implementation. It then generates thin adapters that map that core logic to the unique triggers, tool-calling conventions, and metadata requirements of your target agent environments.

Problem it solves

Writing high-quality AI prompts and agent instructions is time-consuming. However, a skill written for Cursor (.mdc) doesn't natively work in Codex (SKILL.md) or GitHub Copilot instructions. This tool eliminates the manual labor of rewriting rules for every new AI editor, ensuring your agent logic remains consistent and maintainable across your entire dev stack.

Supported Environments

• Codex/OpenAI: Full support for SKILL.md and openai.yaml metadata.
• Claude Code: Intent-driven instructions with clear stop conditions.
• Cursor: Rules (.mdc) with support for auto-attach globs.
• GitHub Copilot: Custom instruction files with applyTo triggers.
• Grok & Gemini CLI: Optimized for CLI-driven loops and long-context anchors.

Why use this skill?

Instead of manually copy-pasting and guessing how to translate tool names (e.g., from Cursor's "Apply Patch" to Codex's terminal commands), this skill uses a structured Portability Principle. It identifies exactly which parts of your skill are platform-agnostic and which require specialized adapters, preventing "hallucinated" tool calls in new environments.