setup-datamodel
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Prompt Injection Surface: The skill processes external data which could contain unexpected instructions. \n
- Ingestion points: The skill ingests user-provided ER diagrams (images, Mermaid text, or descriptions) and site source code (analyzed by a subagent) in Phase 2 and 3 to define the data model. \n
- Boundary markers: A human-in-the-loop approval gate is implemented in Phase 4.2 using the AskUserQuestion tool, requiring explicit user confirmation before any schema modifications occur. \n
- Capability inventory: The skill possesses capabilities to execute shell commands (running scripts via the Bash tool) and perform network operations (interacting with the Dataverse OData API) to manage environment configurations. \n
- Sanitization: Input is parsed into structured formats for presentation; however, the primary safety mechanism is the manual review and approval step by the user. \n- Command Execution: The skill utilizes the Bash tool to execute several local Node.js utility scripts (e.g., dataverse-request.js, verify-dataverse-access.js) for authentication and API communication. This is a standard pattern for skills that automate administrative tasks within a developer environment.
Audit Metadata