The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the browser environment (via the window.__game object) and a project-specific file (game-tests.json) to drive agent logic and reporting.
Ingestion points: Data enters the agent's context through the reading of the game-tests.json file and the inspection of the Three.js game state.
Boundary markers: The instructions do not define boundary markers or safety instructions to prevent the agent from obeying instructions potentially hidden within the game state or the JSON file.
Capability inventory: The skill utilizes the interact tool for simulated input and the test tool for verification and reporting.
Sanitization: No sanitization or validation logic is mentioned for the external content before it influences agent output.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent/user to install and run the paprika3d package using npx -y paprika3d serve, which involves downloading code from an external registry.
[COMMAND_EXECUTION]: The skill relies on external CLI tools (npx, paprika3d) to perform its primary function, executing shell commands to serve the test environment.

threejs-test