enable-tables-offline
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Local Script Execution: The skill executes several Node.js scripts located within the project's internal directories, such as resolve-environment.js and dataverse-request.js. These scripts manage environment configuration and perform API requests. While standard for automation, this pattern relies on the integrity of the local script files provided within the project.
- Remote API Interaction: The skill communicates with the Dataverse API to read and update table metadata. These network operations are necessary to configure the mobile offline profile but involve interacting with external service endpoints. The skill manages this by using dedicated scripts that handle authentication and error reporting.
- Indirect Prompt Injection Surface: The skill processes external data, presenting a potential injection surface. 1. Ingestion points: Data is read from power.config.json, .datamodel-manifest.json, and table metadata from the Dataverse API. 2. Boundary markers: The skill lacks explicit text delimiters for this data but utilizes a mandatory human-in-the-loop plan mode to gate actions. 3. Capability inventory: The skill can execute shell commands via Bash and perform network requests through local Node.js scripts. 4. Sanitization: Data is processed by dedicated scripts rather than direct interpolation into shell commands, which provides a layer of validation.
- Human-in-the-loop Verification: A security best practice is implemented through the use of a plan mode gate. The agent is required to present all intended metadata changes to the user and await explicit confirmation before proceeding with any API write operations.
Audit Metadata