delta-sharing

Pass

Audited by Gen Agent Trust Hub on Jun 28, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a potential surface for indirect prompt injection or SQL injection through the use of string interpolation in Spark SQL queries in SKILL.md and templates/usage_monitoring.py.
  • Ingestion points: Parameters such as share_name, recipient_name, and tenant_id are used in SQL queries, and configuration data is loaded from YAML files via ShareConfigManager.load_config.
  • Boundary markers: None identified; queries do not use delimiters or instructions to disregard embedded content.
  • Capability inventory: The skill performs SQL execution via spark.sql() and administrative tasks (creating shares, recipients, and grants) using the Databricks WorkspaceClient.
  • Sanitization: No input validation or parameterization is observed for the SQL queries or configuration values.
  • [COMMAND_EXECUTION]: The skill uses the pathlib module to perform file system operations. It creates a ~/.delta-sharing/ directory and writes activation URLs to files with restrictive permissions (0o600). While this effectively prevents token exposure in logs, it constitutes direct file system interaction.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 28, 2026, 05:51 PM
Security Audit — agent-trust-hub — delta-sharing