The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the coremltools Python package, which is the official library provided by Apple for CoreML model conversion and optimization.
[COMMAND_EXECUTION]: The skill includes code snippets and workflow templates that involve executing shell commands such as pip install for dependency management and xcodebuild for running performance test suites on iOS/macOS devices.
[INDIRECT_PROMPT_INJECTION]: The skill contains logic for processing external machine learning models (.mlpackage files) and calibration datasets which represent an indirect prompt injection surface.
Ingestion points: Model loading via ct.models.MLModel("Model.mlpackage") and calibration data ingestion in functions like load_calibration_data found in references/compression-techniques.md.
Boundary markers: None; the skill does not explicitly define delimiters to separate model metadata or data samples from system instructions.
Capability inventory: The skill possesses the ability to read and write files (mlmodel.save), execute model inference (model.prediction), and trigger performance benchmarks via shell execution.
Sanitization: The instructions do not specify sanitization or integrity checks for the ingested models or data, relying on the underlying coremltools library for parsing.

coreml-optimizer