The Agent Skills Directory

[SAFE]: The skill provides comprehensive and structured instructions for developing Three.js games, including guidance on performance, state management, and asset handling. The agent is encouraged to use best practices such as resource disposal and avoiding secrets in client code.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it processes untrusted user input.
Ingestion points: User-provided game briefs and existing Three.js project source code (referenced in SKILL.md).
Boundary markers: The instructions include a 'Choose the operating mode' section and a 'Quality bar' to focus the agent, though explicit delimiters for user input are not defined in the provided files.
Capability inventory: The agent has the ability to create and modify files, run build commands, and execute project tests (workflow defined in SKILL.md).
Sanitization: The skill includes instructions to 'profile and play the current build before changing it' and to 'preserve local modifications,' providing a form of behavioral validation against unexpected changes.
[COMMAND_EXECUTION]: The agent is instructed to run build and playtest commands within the user's project workspace to verify the game loop and performance. This behavior is consistent with the skill's primary purpose of delivering a verified, playable game.

threejs-game-generator-skill