The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains instructions to execute sudo apt install meshlab admesh to set up the environment. The use of sudo involves privilege escalation to install system-level packages, which is a potential security risk if the command or package names are tampered with.
[PROMPT_INJECTION]: The skill ingests and processes external 3D model files (STL, 3MF, OBJ) which can contain metadata or complex geometry that could potentially be used for indirect prompt injection to influence the agent's reasoning during the verification and slicing steps.
Ingestion points: Files provided via the source_model input path are read and processed by command-line utilities.
Boundary markers: The instructions do not define clear boundaries or 'ignore' directives for the content within the 3D model files.
Capability inventory: The skill uses Bash to run admesh, grep, and head on the model data and generated G-code.
Sanitization: There is no evidence of sanitization or validation of the model file contents or metadata before they are processed by the toolchain.

prepare-print-model