The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file contains several !cat commands used to dynamically load protocol files from the skills/_shared/protocols/ directory into the agent's context. These commands are executed at skill load time to provide standardized guidelines (e.g., UX protocols, input validation). The paths are hardcoded to internal shared directories and do not involve user-supplied data, making them safe.
[COMMAND_EXECUTION]: The LITE.md file suggests using standard shell commands like ls Assets/Prefabs for environment inspection. This is routine behavior for a development-oriented skill and is presented as a suggested investigative step for the agent rather than a malicious auto-executing command.

3d-spatial-engineer