The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were identified in the skill's instructions or code templates. The skill adheres to best practices for MLOps and infrastructure management.
[CREDENTIALS_UNSAFE]: The skill exhibits a strong security posture by explicitly instructing users to store credentials in secrets managers and specifically forbidding hardcoding credentials in pipeline code.
[INDIRECT_PROMPT_INJECTION]: The skill provides a surface for processing external data (Parquet, CSV, and cloud storage sources), which is standard for ML pipelines. It mitigates potential risks through several layers of defense:
Ingestion points: Data enters the context via ingestion components in references/pipeline-orchestration.md and references/feature-engineering.md.
Boundary markers: The skill uses strict schema validation and Great Expectations suites to define data boundaries and quality gates.
Capability inventory: The skill possesses the capability to write to the file system (joblib.dump, to_parquet) and perform network logging to experiment trackers (mlflow, wandb), which are documented and intended behaviors.
Sanitization: Implements robust sanitization via schema validation, null checks, value range enforcement, and distribution drift detection to ensure data integrity before processing.

ml-pipeline