The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides templates and shell scripts that execute privileged system commands, including the use of sudo for managing firewalls (ufw), network configurations (netplan), and system services (systemctl). Additionally, the TopicWaiter class in SKILL.md uses importlib.import_module for dynamic loading of message types at runtime.
[REMOTE_CODE_EXECUTION]: The deploy-to-robot.sh and ssh-ros2-tunnel.sh scripts in SKILL.md implement remote command execution patterns using ssh and rsync to manage and update software on remote robot hardware.
[PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection through the ingestion of external telemetry and configuration data.
Ingestion points: Data enters the context via DiagnosticArray and Bool topic subscriptions in watchdog_node.py and heartbeat_monitor.py, as well as through user-defined parameters in bringup.launch.py (all within SKILL.md).
Boundary markers: Absent; the skill does not implement delimiters or safety instructions to distinguish data from instructions in these streams.
Capability inventory: The skill scripts in SKILL.md have the capability to execute shell commands, modify system configuration files, and send control signals to robot actuators.
Sanitization: Absent; the skill performs logical checks on data values but does not sanitize string content before logging or processing.
[SAFE]: The skill references established services, such as Google Public DNS (8.8.8.8), within its network configuration templates.

robot-bringup