data-model-creation

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user business requirements to generate Mermaid class diagrams and execute data modeling actions through the modifyDataModel tool.
  • Ingestion points: User-provided entity descriptions, relationships, and business logic specified in the activation contract (SKILL.md).
  • Boundary markers: None specified to delimit user content or warn the agent against instructions embedded within schema descriptions.
  • Capability inventory: The skill utilizes manageDataModel and modifyDataModel to list, read, and create data model structures (SKILL.md).
  • Sanitization: The skill mandates converting descriptions into specific identifiers (PascalCase/camelCase), which acts as a basic structural filter, but no explicit sanitization for malicious instructions is present.
  • [SAFE]: The instructions explicitly forbid the agent from performing HTTP fetches of remote skills or protocols, directing the agent to use local sibling skills only. This is a strong defense against remote resource injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 12:29 AM
Security Audit — agent-trust-hub — data-model-creation