elbelicojackson-hue/haking-code-

Haking Code is an AI-powered terminal agent for cybersecurity professionals, written in TypeScript and running on the Bun runtime. The README is written almost entirely in Chinese. The tool is designed to assist with penetration testing tasks by combining an AI reasoning layer with more than 70 integrated security tools. The project's main selling point is what it calls a four-chain adversarial consensus engine, reached via a slash command called /arena. Instead of one AI answering a question, four separate AI roles argue with each other: one proposes an analysis, one attacks the logic and looks for flaws, one searches the web to verify facts, and one synthesizes the results into a final answer. This is meant to reduce hallucinated or incorrect security conclusions. A second core feature is a hypothesis-driven reverse engineering module. Rather than running tools blindly, it generates a hypothesis about an unknown binary file, picks a tool to test that hypothesis, evaluates the result, and repeats the cycle until all hypotheses are confirmed or a budget is exhausted. The tool also integrates the SecLists password and fuzzing wordlist collection, a FuzzTag payload generation engine borrowed from a Chinese security platform, Shodan reconnaissance, CVE databases tied to NVD and CISA, and a local knowledge base covering phishing techniques, command-and-control frameworks, and antivirus evasion methods. A visual knowledge graph browser with a cyberpunk aesthetic is included for exploring collected security research articles. A Tauri desktop companion app called the Dynamic Island floats at the top of the screen to launch and switch between AI agent sessions. The README documents many recent additions including cross-agent communication via a shared SQLite database and session handoff commands to resume work after closing a terminal.