The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands and remote SSH commands to manage firewall services. The service control script (scripts/service-control.sh) uses an internal mapping and case-based validation for service names and actions, which effectively prevents arbitrary command injection from user input.
[EXTERNAL_DOWNLOADS]: The skill uses curl in its API and backup scripts to interact with OPNsense endpoints. All network operations are directed to the user-provided firewall host and port, following HTTPS protocols with certificate validation enabled by default.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection through the ingestion of external data from the firewall device.
Ingestion points: JSON responses from the OPNsense API (scripts/opnsense-api.sh) and stdout/stderr from service status and log commands (scripts/service-control.sh).
Boundary markers: Absent; firewall responses are provided to the agent without explicit delimiters.
Capability inventory: The skill can execute SSH commands as root, perform system reboots, and modify firewall configurations via the API.
Sanitization: Data is passed to the agent without additional escaping or content filtering.

opnsense-admin