The Agent Skills Directory

[SAFE]: The skill provides instructional content and configuration templates for managing Apache Iceberg tables. All identified behaviors are consistent with its documented purpose of enabling data interoperability.
[CREDENTIALS_UNSAFE]: The skill documentation correctly uses descriptive placeholders (e.g., <pat-token>, <databricks-pat-token>, <service-principal-secret>, <password>) for all sensitive configuration fields. No hardcoded secrets, API keys, or private tokens were found.
[EXTERNAL_DOWNLOADS]: The skill recommends the installation of official, well-known packages including pyiceberg, pyarrow, and adlfs from standard registries. These libraries are necessary for the skill's primary function and originate from trusted sources.
[PROMPT_INJECTION]: The skill operates on external data through Unity Catalog and foreign catalog connections. While this presents a surface for indirect prompt injection (Category 8) from external data sources, the risk is mitigated by the platform's native security model (Unity Catalog RBAC and EXTERNAL USE SCHEMA grants). No instructions to override or bypass agent safety filters were detected.
Ingestion points: Data is ingested via external catalog connections (Snowflake) and PyIceberg/Spark scan operations (4-snowflake-interop.md, 5-external-engine-interop.md).
Boundary markers: Standard SQL delimiters are used; specific instructions regarding embedded data content are absent as it is a data-management tool.
Capability inventory: The skill includes SQL capabilities for DDL/DML including CREATE, MERGE, and DELETE (1-managed-iceberg-tables.md).
Sanitization: Relies on the host platform's Unity Catalog authentication and authorization framework.

databricks-iceberg