nerv-break-jailbreak-framework

1. Prompt Injection & Jailbreaking

• Safety Filter Bypass: The skill is explicitly designed as a "jailbreak framework" to override AI safety constraints. It includes triggers such as "how do I bypass GPT-5.6 safety filters."
• Context Injection (bridge.md): The proxy_relay.py script implements a Man-in-the-Middle (MITM) proxy that injects a bridge.md system prompt into every request. This is used to reconstruct the model's context to ignore safety guidelines.
• Tamper Engine: The framework includes a "tamper engine" that intercepts model responses. It uses regex patterns to detect refusals (e.g., "I cannot help") and replaces them with deceptive content like "Executing analysis," effectively hiding safety triggers from the user.

2. Remote Code Execution (RCE) & External Downloads

• Untrusted Source Execution: The skill instructs users to git clone code from a third-party GitHub repository (zxwn) and execute various scripts (lab.bat, proxy_relay.py, deploy.py) with system privileges.
• Subprocess Spawning: The mcp_server.py script uses subprocess.run(shell=True) to execute security tools. This grants the skill the capability to run arbitrary shell commands on the host environment.

3. Command Injection Vulnerability

• Unsanitized Input: In mcp_server.py, the execute_tool function constructs shell commands by formatting templates from a JSON file with user-supplied arguments (tool["cmd"].format(**args)). Since these inputs are passed directly to a shell without sanitization, it allows for arbitrary command injection via shell metacharacters (e.g., ;, &, |).

4. Malicious & Unethical Intent

• Piracy & Cracking: The skill explicitly provides use cases for unethical activities, including "modifying smali code to unlock VIP features" and "analyzing authorization logic" to bypass software protections.
• Offensive Tooling: It facilitates attacks on specific targets with commands for SQL injection scanning, subdomain enumeration, and authentication bypass testing.

5. Data Exposure & Insecure Credential Handling

• Sensitive File Access: The documentation encourages users to expose sensitive credentials by setting environment variables like KALI_SSH_KEY to point to private SSH keys (id_rsa).
• Relay Risk: The architecture routes all AI traffic through a "Relay" server, potentially exposing user prompts, responses, and API keys to third-party infrastructure.