The Agent Skills Directory

[SAFE]: A detailed analysis of the instruction files and Python scripts confirmed that the skill does not contain any malicious patterns such as prompt injection, data exfiltration, or persistence mechanisms.
[COMMAND_EXECUTION]: The skill utilizes Maya's internal command systems (maya.cmds, mel.eval) and the sibling script mx_run.py to perform 3D modeling tasks. These operations are well-scoped to geometric manipulation and scene hygiene, adhering to standard technical artist practices.
[DATA_EXPOSURE_AND_EXFILTRATION]: No attempts to access sensitive system files (e.g., SSH keys, environment variables) or hardcoded credentials were found. Network activity is limited to documentation references and platform-standard installation commands provided as guidance to the user.
[INDIRECT_PROMPT_INJECTION]: The skill ingests input data ('briefs') from other skills and processes 3D scene data. Although this creates a theoretical attack surface for indirect prompt injection, the operations performed on this data are primarily geometric and deterministic, offering no significant path to bypass agent safety guidelines.

scenario-maya-modeling