dataverse-overview
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to handle and process metadata and records from Microsoft Dataverse environments, which constitute external data sources. Malicious content within these environments could potentially influence agent behavior if not properly delimited.\n
- Ingestion points: Accesses Dataverse table records, schema metadata, and solution components through the SDK and CLI tools.\n
- Boundary markers: The instructions mandate environment and solution confirmation steps by the user, although explicit prompt delimiters for the ingested data are not specified in the overview.\n
- Capability inventory: The skill possesses capabilities for file system modification (solution unpacking), network communication (Dataverse Web API/SDK requests), and command execution via various CLI tools.\n
- Sanitization: There are no explicit data sanitization or escaping protocols described for the external content processed.\n- [COMMAND_EXECUTION]: The skill directs the agent to utilize multiple command-line interfaces, including
dataverse,pac,az, andgh, to perform authentication, environment configuration, and solution management. These tools perform system-level actions and network requests to authorized services.\n- [DYNAMIC_EXECUTION]: The skill relies on the execution of a workspace-local script,scripts/auth.py, for managing authentication tokens. While this script is provisioned by the relateddataverse-connectskill, it involves the execution of local code that is dynamically generated within the project environment.
Audit Metadata