The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill downloads and executes an installation script from an external GitHub repository (mack-a/v2ray-agent) to deploy proxy services. The instructions specify using wget with the --no-check-certificate flag, which bypasses SSL/TLS verification and increases the risk of man-in-the-middle attacks.
[COMMAND_EXECUTION]: Extensive use of SSH is required to perform various administrative tasks on remote hosts, including service management, firewall modification, and kernel tuning. This provides the agent with a broad command execution surface across multiple machines.
[PRIVILEGE_ESCALATION]: The skill operates with root privileges on remote nodes to perform system-level changes such as modifying sysctl parameters, iptables rules, and managing systemd units. These operations are inherently high-privilege and could be misused to alter the security posture of the VPS nodes.
[DATA_EXFILTRATION]: The skill inventories remote hosts and extracts sensitive proxy configurations, including VLESS UUIDs and private keys, to generate shareable links. While the instructions treat these as sensitive, the process involves transmitting and processing credentials within the agent's context.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and parses configuration files (realm.toml, Xray JSON configs) from remote hosts using shell tools like cat, grep, and sed. These ingestion points lack sanitization or boundary markers, creating a surface where malicious content embedded in remote files could influence the agent's execution logic.
[EXTERNAL_DOWNLOADS]: In addition to the installation script, the skill installs several packages (wireguard-tools, iperf3, etc.) using apt-get on remote hosts, creating a dependency on external package registries.

proxy-networking