doccker/cc-use-exp

What it solves

This project provides a maintainable configuration system for AI coding assistants, eliminating the need to repeatedly define technical stacks, project structures, and coding standards in every session. It aims to reduce common AI failures—such as breaking API contracts or performing dangerous operations—by establishing a structured set of global principles, approval rules, language-specific skills, and task workflows across multiple AI tools.

How it works

The system uses a layered loading approach to minimize context burden while ensuring consistency:

1. Global Rules: Permanent defensive rules (e.g., preventing test tampering or over-engineering) that are always active.
2. Skills: Domain-specific guidelines (e.g., Go, Java, Python, or API safety) that are loaded automatically when relevant files are modified.
3. Workflows/Commands: Explicitly triggered tasks (e.g., /fix, /review, /commit-msg) for high-frequency operations.

It includes synchronization scripts (sync-config.sh, sync-skill.sh) that distribute these configurations from a central source (primarily Claude Code) to various tool-specific directories in the user's home folder, ensuring a unified experience across different IDEs and CLIs.

Who it’s for

Developers who use one or more of the following AI coding tools:

• Claude Code
• Antigravity CLI
• Gemini CLI
• Codex
• Cursor
• GitHub Copilot

Highlights

• Cross-Tool Synchronization: Maintain configurations once and sync them across six different AI assistants.
• Defensive Guardrails: Built-in rules to prevent dangerous commands, excessive refactoring, and API contract drift.
• ** leverages Tool-Specific Mechanisms**: Utilizes .mdc files for Cursor, AGENTS.md for Codex/Antigravity, and plugins for Claude Code.
• Incremental Deployment: Specifically for Codex, it uses managed blocks to merge configurations without overwriting user authentication or history files.