The Agent Skills Directory

[SAFE]: The skill implementation follows industry best practices for networked game mechanics and contains no malicious code or patterns.
[PROMPT_INJECTION]: Analysis of the skill's instructions and documentation shows no attempts to bypass safety filters or manipulate agent behavior.
Ingestion points: Network data is received via socket.onmessage in scripts/building-network-manager.js.
Boundary markers: Data is processed as structured JSON; no specific natural language delimiters are present.
Capability inventory: The skill is restricted to virtual building state management and lacks access to the host filesystem, shell, or arbitrary networking.
Sanitization: Inbound building requests are validated against permission and collision rules in the server logic.
[DATA_EXFILTRATION]: The skill establishes network connections to a specified server for game state updates. It does not access or transmit sensitive credentials or private files.
[EXTERNAL_DOWNLOADS]: The skill utilizes the reputable three library for 3D rendering. No unverified third-party scripts or packages are downloaded or executed.
[COMMAND_EXECUTION]: No scripts within the skill utilize eval(), exec(), or other methods to run shell commands or dynamic code.

multiplayer-building