The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill provides instructions to install a custom Certificate Authority (CA) into the system's root trust store using administrative commands such as sudo (Linux/macOS) and certutil (Windows). This grant allows the proxy to intercept, decrypt, and re-encrypt all HTTPS traffic on the host machine, creating a severe man-in-the-middle (MITM) risk.
[EXTERNAL_DOWNLOADS]: The installation process involves cloning code from an unverified GitHub repository (github.com/masterking32/MasterHttpRelayVPN).
[EXTERNAL_DOWNLOADS]: The documentation recommends using an unofficial PyPI mirror (mirror-pypi.runflare.com) for dependency installation, which could potentially serve malicious or tampered versions of Python packages.
[REMOTE_CODE_EXECUTION]: The skill requires the user to run multiple Python scripts and shell scripts downloaded from the external repository, including main.py, setup.py, and start.sh / start.bat.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to modify system certificate stores and manage local services.
[CREDENTIALS_UNSAFE]: The configuration process involves hardcoding a shared authentication secret (AUTH_KEY) in plaintext within both the Google Apps Script deployment and a local JSON configuration file.
[INDIRECT_PROMPT_INJECTION]: By acting as a proxy for all web traffic, the skill creates a significant ingestion point for untrusted external data. While the current instructions do not show direct interpolation of this data into agent prompts, the architecture constitutes a high-risk surface for indirect injection attacks from malicious websites.

masterhttprelayvpn-proxy