artemnovichkov/xcode-27-system-prompts

What it solves

This repository provides a collection of internal system prompts and documentation extracted from the Xcode 27 beta, revealing how Apple's AI coding assistant is instructed to behave, reason, and integrate code into iOS and macOS projects.

How it works

The project organizes extracted .idechatprompttemplate files and Markdown guides. These include:

• System Prompts: Core instructions for different modes, such as basic coding assistance, complex reasoning, and tool-assisted editing.
• Agent Skills: Specialized prompts paired with reference documentation for specific tasks like auditing security settings, adopting C bounds safety, or implementing App Intents.
• Context Providers: Templates that define how the AI receives information about the current file, selection, or project structure.
• Technical Documentation: Guides on on-device LLMs, Visual Intelligence, and new SwiftUI/UIKit features.

Who it’s for

• AI Prompt Engineers: Those looking to study Apple's professional-grade prompting strategies for coding assistants.
• Apple Platform Developers: Developers wanting to understand the latest beta features and the idiomatic way to implement them using AI.
• LLM Researchers: People studying how integrated IDE AI agents are structured to handle tool-use and context management.

Highlights

• Diverse Prompt Modes: Includes specific prompts for "Reasoning," "Fast Apply," and "Planner/Executor" workflows.
• Agentic Capabilities: Reveals the use of MCP tools, documentation search, and phased plan-and-approve workflows for security audits.
• Apple-First Philosophy: Documents the strict preference for Swift Concurrency, the Swift Testing framework, and official Apple APIs over third-party libraries.
• Code Editing Strategy: Highlights Apple's approach of requiring complete file replacements rather than partial edits to ensure syntactic validity.