The Agent Skills Directory

[SAFE]: The skill implements robust security patterns for multiplayer games, including server-side validation of client actions, rate limiting for RPC calls (net_packet_rate_limiter.gd), and explicit authority checks to prevent cheating or unauthorized state manipulation.
[COMMAND_EXECUTION]: The script net_headless_server_auto_start.gd utilizes OS.get_cmdline_args() to allow configuring the server port via CLI arguments (e.g., --port=7777). This is a standard practice for dedicated server deployments and includes safe integer conversion of the input.
[INDIRECT_PROMPT_INJECTION]: The skill features multiple data ingestion points where it processes untrusted network data, such as UDP discovery packets in net_lan_discovery.gd and RPC dictionaries in server_authoritative_controller.gd.
Ingestion points: udp.get_packet() in net_lan_discovery.gd, send_input RPC in server_authoritative_controller.gd.
Boundary markers: The code uses Godot's high-level multiplayer API which segregates network data from application logic via RPC decorators.
Capability inventory: The skill is limited to game state manipulation (physics, position, health) and does not possess capabilities for file system writing, shell execution, or sensitive data access.
Sanitization: The code utilizes typed variables and explicit type casting (e.g., .to_int(), .to_utf8_buffer()) when processing network buffers, minimizing the risk of exploitation via malformed data.

godot-multiplayer-networking