ontology-stack-builder

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the Bash tool to run internal Python scripts (introspect_schema.py, parse_owl.py, generate_ontology_sql.py) for processing metadata and generating SQL artifacts. This execution is confined to a strictly managed workflow that requires explicit user confirmation for every transition.
  • [EXTERNAL_DOWNLOADS]: Dependencies defined in pyproject.toml consist of well-known, legitimate Python packages including rdflib, pyyaml, streamlit, and streamlit-agraph used for data parsing and visualization.
  • [PROMPT_INJECTION]: Instructions in SKILL.md define 'MANDATORY STOPPING POINT' and 'GLOBAL RULE' constraints. These are designed to ensure the agent does not proceed autonomously, effectively mitigating risks associated with unintended execution or prompt-based overrides.
  • [DATA_EXFILTRATION]: No unauthorized network operations or credential harvesting patterns were detected. The skill follows Snowflake's standard security practices for managing session tokens within containerized services (SPCS).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 11:10 PM
Security Audit — agent-trust-hub — ontology-stack-builder