The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes templates that use kubectl and mlflow to manage infrastructure and job execution. These are standard operations for deploying and training models in cloud and cluster environments.
[DYNAMIC_EXECUTION]: Several Python scripts and templates utilize mlflow.pyfunc.load_model and joblib for model and artifact management. While these involve deserialization, they are core components of the MLOps workflow and the skill provides examples of validating data quality prior to processing.
[EXTERNAL_DOWNLOADS]: The skill references and installs standard packages from official registries (PyPI, Conda-Forge) such as mlflow, xgboost, and scikit-learn. All external references target official documentation or reputable open-source projects.
[INDIRECT_PROMPT_INJECTION]: The skill defines pipelines that process external data files. 1. Ingestion points: templates/kubeflow-pipeline.yaml (S3 parquet files), templates/mlflow-project/validate.py (local CSV files). 2. Boundary markers: Absent, as the skill deals with structured numerical data rather than natural language prompts. 3. Capability inventory: Cluster deployment via kubectl, job execution via mlflow, and file system writes via joblib. 4. Sanitization: The skill incorporates great-expectations to validate data schema and values before processing.

mlops