semantic-model-analyst

Pass

Audited by Gen Agent Trust Hub on May 18, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to capture user requirements and store them in the initial_request field within the generated semantic model. This introduces an attack surface where malicious instructions could be embedded in the initial request and processed by the agent in future sessions. However, the use of YAML literal blocks for storage provides standard data/instruction separation.
  • Ingestion points: The initial_request field defined in references/semantic-model-template.md captures verbatim user input.
  • Boundary markers: The template uses YAML block markers and distinct markdown sections to delimit user-provided content.
  • Capability inventory: The skill utilizes vendor-specific tools (semantius call crud) to perform database operations including creating and updating entities and fields.
  • Sanitization: The skill focuses on structural correctness but does not include explicit instructions for sanitizing or escaping the user's initial prose request.
Audit Metadata
Risk Level
SAFE
Analyzed
May 18, 2026, 11:43 AM
Security Audit — agent-trust-hub — semantic-model-analyst