3641397194-wq/gpt6-Astra

What it solves

This project provides a centralized workbench for modifying and configuring AI coding assistants and LLM-based software (such as Cursor, Claude Code, and Kimi K3) to enable "cracking" (破甲) or custom configuration of their system prompts and agentic behaviors. It simplifies the process of writing specific configuration files and system prompts to various local directories across Windows, macOS, and Linux.

How it works

The tool operates by writing specific markdown files (like SKILL.md or CLAUDE.md) and configuration files (config.toml) to the known configuration directories of various AI tools. It creates backups of original files before modification. Users can apply these changes with a one-click "crack" and revert them using a one-click uninstall feature.

Who it’s for

Developers and power users who want to customize the behavior of their AI coding assistants and LLM tools, or those looking to bypass certain restrictions or apply custom system prompts to these tools.

Highlights

• Multi-platform support: Works across Windows, macOS, and Linux.
• Wide software coverage: Supports 11 different AI tools including GPT-6 Astra, Claude Code, Grok, DeepSeek, GLM, Gemini, Doubao, WorkBuddy, Cursor, MiMo, and Kimi K3.
• Integrated Proxy: Includes a proxy service (coldcoffeeai.com/v1) for managing API keys and configurations.
• IDA Pro Integration: Provides a simplified setup for IDA Pro Chinese localization and connects AI tools to a local IDA instance via MCP (Model Context Protocol).
• Reverse Engineering Tooling: Includes a toolkit for IDA Pro 9.x localization.