azure-logic-apps-standard-cli
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes various local configuration files (e.g., host.json, workflow.json, connections.json) and external metadata (e.g., connector catalogs). While it mandates validation steps, it lacks explicit prompt-level delimiters to prevent potential instructions embedded in these files from influencing agent behavior.\n
- Ingestion points: Reads project and workflow configuration files during authoring, inspection, and result analysis steps.\n
- Boundary markers: Absent; instructions interpolate data from files without specific escaping or boundary markers.\n
- Capability inventory: Includes file system modifications, network requests to Azure services, and local process execution (e.g., starting runtimes, running tests).\n
- Sanitization: Relies on integrated CLI validation commands (az logicapp workflow validate) and mandatory manual review of mutations.\n- [EXTERNAL_DOWNLOADS]: The skill manages the installation of the logicapp Azure CLI extension and its required development tools (Core Tools, Node.js, Azurite) from official and trusted sources.\n
- Evidence: Commands like az extension add and az logicapp dependency install are used to fetch resources from the official Microsoft extension index and standard registries.\n- [DYNAMIC_EXECUTION]: The skill provides the ability to run native .NET workflow tests, which involves executing compiled code. This capability is explicitly protected by requiring the user to provide specific trust and acceptance flags.\n
- Evidence: The az logicapp test run command in references/testing.md requires the --trust-project and --accept-sdk-terms flags to authorize the execution of test code.
Audit Metadata