The Agent Skills Directory

[SAFE]: The skill describes a standard machine learning optimization workflow (post-training quantization) using legitimate libraries such as OpenVINO and NNCF. No security issues were detected in the instructions or code snippets.
[INDIRECT_PROMPT_INJECTION]: The skill provides an interface for processing external model files (.xml) and datasets. While processing untrusted binary or data files is a theoretical attack surface, this is standard functionality for ML tools and no malicious exploitation patterns are present.
Ingestion points: Path to exported OpenVINO model XML and path to dataset directory in create_engine().
Boundary markers: None explicitly defined in the provided snippets.
Capability inventory: Uses the getitune engine to optimize models, validate accuracy with test(), and run inference with predict().
Sanitization: Relies on the security and validation logic within the getitune and OpenVINO libraries.

getitune-optimizing-a-model