The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses multiple bash scripts to execute system utilities including wg, ip, ifconfig, and netstat to monitor VPN tunnel health, handshake timings, and transfer statistics.
[PRIVILEGE_ESCALATION]: Extensive use of ${WG_SUDO:+sudo} gives the agent administrative access to system-level VPN configurations and network interface states, which is necessary for WireGuard management.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and parses untrusted data from the local system, such as the output of wg show and directory listings, which could be exploited if resource names were maliciously crafted.
Ingestion points: Command output from wg show interfaces and directory listings of /etc/wireguard/*.conf (SKILL.md).
Boundary markers: The skill lacks explicit sanitization or delimitation markers for parsing command output.
Capability inventory: Execution of administrative commands via sudo for monitoring and potential management of interfaces (SKILL.md).
Sanitization: Uses basic shell tools like awk and tr for parsing, which does not provide robust validation of interface or peer identifiers.
[CREDENTIALS_UNSAFE]: The skill accesses the sensitive directory /etc/wireguard/, which contains configuration files including private keys. While the safety rules explicitly forbid displaying these keys, the access to this path involves sensitive credential storage locations.

managing-wireguard